University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Chalmers University of Technology | 6 Governing Structure and Management Control for
Water Distribution in South Africa
This chapter will explain the administrate system responsible for South Africa’s water
resources and drinking water production. The most important bodies, organizations,
departments etc. that are involved will be presented and the cooperation between these will be
clarified. One of the most substantial tools for water efficiency and safety is water safety
plans (WSP), and will be further described.
6.1 Department of Water Affairs
The Department of Water Affairs, DWA, earlier a part of Department of Water and Forestry,
DWAF, is the central unit responsible for South Africa’s water affairs. The department has
since the end of the 20th century acted as the custodian for the country’s water resources and
been responsible for development of the water resource infrastructure (DWAF, 1994; DBSA,
2006). The main task for the DWA is to formulate and implement policy documents
concerning South Africa’s water resources and monitor that all South Africans have access to
clean and safe water and sanitation service (DWA, 2011). A definition of the term “safe
drinking water” is presented in the report Drinking Water Quality Management Guide for
Water Services Authorities (DWA, 2005); the report states that safe drinking water is water
that: “…does not pose a significant risk to health over a lifetime of consumption, including
different sensitivities that may occur between life stages…”
In 1994, the DWAF composed a policy document called: Water Supply and Sanitation policy,
or the “White Paper”. The document was intended to, in a provocative way challenge South
Africans involved with water questions at all levels, to participate and contribute towards a
sustainable water and sanitation policy (DWAF, 1994). Later on came the Water and Service
act of 1997 (act no 108, 1997), which declared the rights for every citizen to have access to
basic water supply and sanitation service (DWAF, 1997). The act defines the responsibility
distribution between water service authorities/administrations and water service providers
(DBSA, 2006) and also places a duty on South Africa’s Water Safety Authorities (WSA) to
provide and maintain the current water safety plan (WSP) for their area of jurisdiction (BWM,
2010a). In South African acts and documents WSP’s are often called Water Safety
Development Plans (WSDP), which basically is the same methodology as WHO’s water
safety plan (WSP), see chapter 6.3. In this report the term WSP will be used.
In 2004/2005 DWA presented a regulation program, intended to increase drinking water
quality and facilitate a sustainable drinking water management, called The drinking water
quality regulation program. The reason behind the program was that in 2004, less than 50%
of the WSAs could monitor their drinking water quality according to legislated requirements,
i.e. there existed a widespread lack of knowledge of how to monitor drinking water quality in
a proper way. In the beginning of 2008, three years after the program was implemented, the
monitoring compliance had increased to 100%. However, this does not mean that 100% of the
drinking water quality met the national standard; it only means that all municipalities could
monitor their quality according to legislated requirements. As the actual drinking water
quality still was insufficient, the Blue Drop Certification was presented in the summer of
2009. The Blue Drop certificate takes into account and grades 9 parameters, e.g. the water
safety plan, compliance with SANS 241 standards presented in the report Drinking Water
Quality Management Guide for Water Services Authorities (2005), process control and
maintenance ability etc., to indicate how well municipalities manage their drinking water
supply. To be assigned the Blue Drop certificate is seen as a big acknowledgement in South
20 |
Chalmers University of Technology | Africa and shows that the drinking water system maintains a high international standard
(DWA, 2010). There exists a similar certification for wastewater management, called the
Green Drop Certificate.
6.2 Water Safety Authorities/administrations and Water Boards
A water safety authority, WSA, is e.g. a district municipality or authorized local municipality
that provides water to the inhabitants within its area of jurisdiction (DWAF, 2003). The
DWAF (2003, 2005) state that, the primary legal responsibility for providing safe drinking
water to the consumers rests with the WSA. Also the Service act of 1997 state that: “Every
WSA has a duty to all consumers or potential consumers in its area of jurisdiction to
progressively ensure efficient, affordable, economical and sustainable access to water
services” and further explains that a WSA also can function as the water service provider and
that “No person may operate as a water services provider without the approval of the water
services authority having jurisdiction in the area in question” (DWAF, 1997). The main tasks
for the WSA’s are to:
Supply the inhabitants in its area of jurisdiction with basic water service.
Monitor and evaluate the quality of the drinking water against national standards.
In case of an emerging health risk, communicate it to the consumers and appropriate
authorities.
Local WSA’s can either supply their inhabitants themselves with basic water service, or they
can involve external water service providers, e.g. non-governmental organizations, private
companies, or so called water boards. Water boards are state owned organization/entities that
operate and handle dams, wastewater systems, water supply infrastructure etc. Their task is to
work as water utilities and, in cooperation with WSAs, provide people with basic water
service. Today’s water boards were originally private owned companies or organizations that
saw the emerging need for water supply as a business opportunity. This form of ownership,
and the lack of control, resulted in high tariffs and misuse, which forced the authorities to
legislate the area of water production and supply. The first act was written in 1956 and was
later followed by the Water Service Act of 1997, which brought all different water boards
under its sphere and gave the Department of Water Affairs the option to establish and resolve
water boards. It also defines water boards as public entities, with control and shareholding by
the national government. (DBSA, 2006)
6.3 Water Safety Plan
In 2004 the World Health Organization, WHO, presented the WSP framework as an initiative
to facilitate the process of risk assessment and risk management connected to water (WHO,
2004). The framework was required, since there was a need for an increase in both awareness
and understanding of risk issues concerning drinking water (Rosén et al., 2007).
The main purpose of the WSP is to provide guidance to be able to produce sustainable and
safe drinking water (WHO, 2005). A comprehensive WSP should include the whole supply
chain, i.e. a source to tap approach. The use of a multi-barrier approach implies that actions
are taken at all levels in the chain to assure that safe drinking water is delivered to the
consumer. According to WHO (2005) the main objectives of WSP’s are to ensure:
Raw water sources do not get contaminated or that the raw water supply is interrupted
in any other way.
21 |
Chalmers University of Technology | 7 General Description over Beaufort West Municipality
Beaufort West Municipality (Figure 7.1) is situated in central Karoo, one of the driest areas in
South Africa, and functions as the economic, political and administrative centre of the central
Karoo. There are 41 000 estimated inhabitants in Beaufort West municipality and the
municipality consists of three towns; Beaufort West, Merweville and Nelspoort, where
Beaufort West is the administrative centre. The municipality functions as both drinking water
authority and drinking water provider and the three society’s drinking water supply systems
are independent of each other and supported by raw water from different sources (DWA,
2010).
Figure 7.1 Map over South Africa (Welt-Atlas, 2011).
The highway N1 passes through Beaufort West and is one of the major driving sources for
economic growth, while agriculture and agri-processing forms the backbone of Beaufort west
economy. Agriculture accounts for the largest labour force of the population (BWM, 2010a).
Next to Beaufort West lies one of South Africa’s largest national parks, Karoo, which attracts
thousands of tourists each year.
7.1 Water Supply System in Beaufort West (without WRP)
Beaufort West relies on surface water from the Gamka Dam and groundwater from several
boreholes spread widely around the town. The water from the dam is treated at a local WTP
with the treatment steps: flocculation, stabilization, filtration and chlorination. The treated
water from the dam is then mixed with the borehole groundwater and via three reservoirs
distributed to the consumers (Figure 7.2).
23 |
Chalmers University of Technology | Figure 7.2 Schematic layout over Beaufort West’s drinking water supply system before construction of
WRP (BWM, 2010a).
Beaufort West also has a sewage system connected to the local WWTP, for all formal
households, while storm water is not directed to the WWTP. No larger industries are situated
in the area meaning that it mainly is wastewater from households that is being treated in the
WWTP. For more information on the WWTP, see chapter 9.1. (BWM. 2010a).
7.2 Tariffs
The tariff for households in Beaufort West is defined by using a rising block tariff structure
(Table 7.1). The first block, Free Basic Water, corresponds to a monthly water quantity of
6000 l per household or 25 l/person per day, also defined in the basic water service. This
amount of water is provided free for consumers who qualify for indigent relief. As
comparison, the mean consumption of water in households in Sweden is 160 l/person per day
(Svenskt Vatten, 2011). The intention of the rising tariff structure is to discourage wasteful or
inefficient use of water, and punitive tariffs have been introduced for excessive water
consumption. The tariffs are also influenced by different drought phases meaning that it costs
more to consume water during a severe drought compared to during normal conditions. Basic
water service is however excluded from this cost-increase since basic water service should be
available for people in spite of their income (BWM, 2010a).
24 |
Chalmers University of Technology | Table 7.1 Block tariff structure in Beaufort West Municipality (BWM, 2010a).
Block Normal Drought Drought Drought
Comment
(kl/month) condition Phase 1 Phase 2 Phase 3
Free Basic Water
0-6 R3-64 R3-64 R3-64 R3-64
Low volume use.
7-10 R4-17 R4-17
R4-17
11-15 R4-17 R5-46 R8-34
Typical use volume, including
16-20 R6-73 R10-43
garden irrigation.
R5-46
21-25
R8-92 R12-51
26-30 R4-58
R6-73
Above average use, including
31-50
garden irrigation.
51-60
R12-74 R14-60
Wasteful use and/or severe garden
61-100 R4-96 R8-92
irrigation.
Significant waste and/or
>100
unnecessary garden irrigation.
The step block tariff structure has not been implemented for industrial and commercial
consumers, and there is no system to charge industries for effluent that needs to be
extraordinary treated. However, since there are no major industries in Beaufort West it is not a
big issue.
At present, 1 m3 of drinking water costs approximately 0.9 South African Rand to purify with
the conventional system (BWM, 2011a). Treating water with a reclamation system, explained
further in chapter 9.2, costs approximately double that amount4 and treating seawater with
desalination plants costs almost four times as much5.
7.3 Beaufort West Municipality’s Water Safety Plan 2010/2011
BWM constructed their first WSP in 2010/2011 (BWM, 2010a). The result from the report
shows that the water services provided for the inhabitants in the municipality generally meet
national standards according to standard SANS 241 presented in the report Drinking Water
Quality Management Guide for Water Services Authorities (2005). For more detailed
information regarding the municipality’s water quality, see the report - Annual publication of
drinking water quality performance against SANS 241, published at
http://www.beaufortwestmun.co.za/ (BWM, 2010b).
4 Christopher Wright, Beuafort West Mun. Manager: Technical Services 2011-04-21 (Personal communication)
5 Cobus Oliver, Veolia Water South Africa, Engineering Manager, 2011-05-23 (PP presentation)
25 |
Chalmers University of Technology | 7.3.1 Essential Shortcomings
The municipality is today providing help to farms that do not have basic sanitation service to
install so called ventilated improved pits (VIP). There are economical resource limitations that
do not make it possible to provide waterborne sanitation systems for the moment. The future
goal is to investigate and determine the quantity and quality standard of available drinking
water and to be able to provide all farms in the municipality with drinking water that meets
national standards. (BWM, 2010a)
There are problems with informal settlements and backyard dwellers in the municipality and
there exists a backlog i.e. a shortage, to support these inhabitants with in-house water
connections and sanitation service. The goal is that all houses in informal areas shall be
provided with basic water services that meet national standards. Today many households are
using shared services, i.e. shared water taps and toilets between households/families.
According to BWM (2010a) it is not a permanent solution and the maintenance cost of these
shared services is not financially sustainable.
As an addition to these problems there is also a large backlog of houses, approximately 3000,
despite that 1500 new houses were built between 2004 and 2009. Hence, large future
investments are required to supply new and old households with sufficient drinking water and
sanitary service. (BWM, 2010a)
7.4 Blue/Green Drop Certificate
Beaufort West was issued with the Blue Drop certificate in 2010 (DWA, 2010). The town’s
water supply system scored 95% compliance towards the nine evaluated performance areas.
However, that is also the minimum limit to be awarded the certificate so there exist
possibilities for improvements. The 2010 Blue Drop report (DWA, 2010) further states that
“…Beaufort West Local Municipality displayed impressive improvement since the 2009
assessment…”.
The municipality has applied for Green Drop certification and this is still under investigation.
The municipality expects an answer about their score during the end of 2011. Fulfilling Green
Drop requirements is an important aspect in future water production since it would certify that
their treated wastewater, which is the raw water source of the Reclamation Plant, will have a
good quality.
7.5 Water Shortage - Construction of a Reclamtion Plant
In 2010 a long drought was severely affecting the drinking water production of Beaufort
West. The Gamka dam, which is the main raw water source for Beaufort West, was empty by
the end of 2010, resulting in a lack of raw water for the drinking water production. Water
from boreholes was not enough and drinking water had to be transported to the town by trucks
to fulfill the immediate need. The drought was the worst drought during the last century, and
made the municipality aware of the extent of the problem.
To encounter this severe situation several options were investigated and implemented, e.g.
managing water losses, optimize existing aquifers and exploring new groundwater sources.
However, none of these options were sufficient, resulting that the construction of a
Wastewater Reclamation Plant (WRP) was decided on as the only suitable solution. The WRP
will increase the drinking water production and make it possible to supply the present and
growing population with drinking water that fulfills quantity and quality standards. The WRP
26 |
Chalmers University of Technology | will produce water all year around, and thereby relieving pressure on the Gamka dam and
boreholes, so that there is storage of water if a similar drought occurs.
Beaufort West’s reclamation system is the first direct WRP producing drinking water in South
Africa. The plant was initially intended to be constructed as a public-private partnership
(PPP), i.e. financed and operated by an external contractor. However, the municipality of
Beaufort West was assigned a governmental grant from the drought relief fund, resulting
therein that the municipality could finance the plant. A tender document specified
requirements for the project and the plant was built according to the design and build
approach6. The contractor designed the plant in Beaufort West, after a multi-barrier concept
successfully used at the New Goreangab reclamation plan. The following barriers are used in
Beaufort West: Ferric-chloride dosing at inlet to the secondary settling of the WWTP, Pre-
chlorination, Sedimentation basin, Post-chlorination, Rapid sand filtration, Ultra filtration,
Reverse Osmosis, UV-Hydrogen peroxide, Final chlorination. The contractor is also
responsible for operating the plant for 20 years. For more information about Beaufort West’s
Reclamation Plant, see chapter 9.2.
Reclamation systems tend to be complex and connected to higher risks since the raw water
contains more pathogens, which requires more technically advanced barriers, compared to
conventional water treatment plant. Consequently, a comprehensive risk assessment is
required to identify hazards and estimate risk levels. It is also crucial to have a dialog and
acceptance from the inhabitants since doubts towards drinking water produced from enhanced
wastewater is inevitable. Furthermore this type of systems is expected to become more
common in South Africa as well as other countries suffering from water scarcity. Therefore
more knowledge in the field is crucial for successful continuing progress and development.
The risk assessment for the Reclamation Plant was performed as a case study and is presented
in chapter 10.
6 Christopher Wright, Beuafort West Mun. Manager: Technical Services 2011-04-19 (Personal communication)
27 |
Chalmers University of Technology | 8 Reclamation System in Windhoek, Namibia
The Reclamation Plant in Windhoek, New Goreangab, has reclaimed wastewater and treated
it to potable water for almost 60 years. It has for many years been the only direct Wastewater
Reclamation Plant (WRP) in the world and a lot of experience has been gathered during
operation. New barriers have been introduced and problems had to be solved as they occurred.
Since it was the first of its kind the WRP has also worked as a research project to gather
information and expertise in the field of reclaiming wastewater and treating it to drinking
water. Due to the research approach, additional extensive funding has also been available,
which is not the case for Beaufort West.
The WRP in Beaufort West is based on a multi-barrier concept that has successfully been
used in Windhoek. The two WRP’s share many solutions and difficulties. This is why a case
study was performed in Windhoek, 2011. Both the WRP in Goreangab and Gammams
Wastewater Treatment Plant (WWTP) were studied since they are equally important to
achieve a good result. The following results came up during discussions with Jurgen Menge,
analysis responsible for Gammams WWTP City of Windhoek, John Esterhuizen, General
Manager for WINGOC Water Reclamation Plant and Truddy Theron-Beukes, consultant and
former employee at City of Windhoek.
In the long run, indirect risks have been harder to solve and a good contract between
the WWTP and the WRP is crucial.
The reclamation system in Windhoek would not have been constructed in the same
way if designed today. RO would have been used instead. Partially since the
membranes and the technique overall has become cheaper, but also since the salt
content in the treated water from the WRP is too high, why reverse osmosis is
evaluated. Brine may become a problem in arid parts, and the dilution factor is
important to consider.
Politicians need to be involved to a large degree to make adequate decisions and
investments. WRP’s does not consist of a one-time investment but needs to be
maintained and future investment needs to be accounted for.
The majority of the long-term disputes and indirect risks have occurred due to the fact
that the WWTP and the WRP are operated by different owners. Optimally, the system
is operated and owned by the same owner to be able to make correct and most cost-
efficient investments and adjustments.
J. Esterhuizen considers that reverse osmosis is the best solution since it requires less
experience to operate and is safer.
Monitoring becomes substantially more comprehensive for WRP’s compared to using
surface water or groundwater. More monitoring means higher costs and it also requires
highly skilled personal.
On high-tech plants maintenance and availability of spare parts cannot be
compromised. Gammans wastewater treatment plant indicated this as a problem.
When they have needed to order parts it has taken them much time due to slow
bureaucracy and not always easy to get approval. This is not indicated as a problem
for Goreangab since they are operated as a private company and do not have the same
problem with slow bureaucracy.
The water from the Reclamation Plant should have a higher, or at least as high, quality
as the water from the conventional system.
28 |
Chalmers University of Technology | A specific flux for the UF and RO should be incorporated in the contract to avoid the
operator of the WRP operating the membranes under too high pressure and thereby
reducing the life span of the membranes.
Public acceptance is crucial for a successful project. Therefore it becomes important to
have a well-adapted information campaign and dialogue with the consumers. It has
sometimes proven to be a challenge to decide how much information should be
released to avoid external threats as sabotage, terrorist attacks etc. There is also a
widespread resistance towards drinking enhanced wastewater among Muslim people,
which needs to be considered.
Human errors are very difficult to control. To avoid this as much as possible staff
needs continuous training by regular programs and external audits. If this is not done
the whole project may slowly deteriorate into failure. A good, and proven, way to
minimize human errors is to get an ISO-certification, where external audits certifies
that staff and operative responsibilities can operate the plant and handle deviations in a
appropriate way.
29 |
Chalmers University of Technology | 9 Beaufort West Reclamation System Description
The reclamation system in Beaufort West uses wastewater as its only raw water source to
produce drinking water. The system consists of an existing WWTP with conventional
treatment and a new constructed membrane Wastewater Reclamation Plant (WRP). The
system is a direct reclamation system, which means that it, compared to indirect reclamation
system, cannot benefit from any dilution of the wastewater due to mixing with other water
sources. Wastewater is a very complex raw water source, compared to groundwater or surface
water, since it contains high amount of pathogens and the quality and quantity has a tendency
to vary. There are also compounds which effect to humans in a longer perspective is
unknown. The sewage system is separated, meaning that no storm water is supposed to enter
the sewage system; further no industrial effluent is diverted to the WWTP. Still the flow to the
WWTP is affected during heavy rain7.
The WRP in Beaufort West uses fewer barriers compared to the WRP in Windhoek. The
treatment process is therefore easier to operate since it mainly relies on two membrane
filtration barriers, ultra filtration and reverse osmosis, which are highly atomized. WINGOC
stated that if their plant had been built today, they would also have been using reverse osmosis
due to that it is safer and easier to operate8. A system like Beaufort West’s that is highly
automated and using reverse osmosis with connected alarms does not require as much skilled
personal for the daily handling, as the system in Windhoek.
Due to advanced and expensive treatment barriers Beaufort West municipality could not carry
the installation cost alone. The WRP was granted funding from the government, due to the
extreme water shortage, and is today owned by the municipality. The same contractor that has
constructed the plant is also responsible for the daily operation as well as the maintenance
work on a 20 years contract period. Production rate will start at a minimum of 1 Ml per day,
with an increase of 10% over a period of ten years. This means that after the first ten years of
operation, when reaching design capacity, the plant needs to produce water for 20 hours per
day. The contractor does have mandate to change operation or demand additional barriers in
the WWTP if considered necessary with regard to the reclamation process. The WWTP is
both owned and operated by the municipality.
9.1 Treatement Barriers – Wastewater Treatment Plant
The Wastewater Treatment Plant, WWTP, in Beaufort West is a rather uncomplicated system
and relies on conventional treatment techniques. The plant has two parallel treatment trains
that use different treatment techniques, but the WRP uses only one (Figure 9.1). The reason
for that only one of the two treatment trains is used is because one is enough to support the
WRP with raw water and that treatment train is also more efficient.
7 Christopher Wright, Beuafort West Mun. Manager: Technical Services 2011-04-19 (Personal communication)
8 John Esterhuizen, General Manager: WINGOC 2011-03-29 (Personal communication)
30 |
Chalmers University of Technology | Figure 9.1 Conceptual model over Beaufort West WWTP.
Very few changes have been made to the WWTP after the introduction of the WRP. On
initiative of the contractor for the WRP, ferric chloride is added after the activated sludge
process to increase settling and to have a more efficient removal of phosphates. This has
shown to be a big improvement for the reclamation system.
9.1.1 Screening and Grit Removal
The screening and grit removal is the first treatment step. A new screener was recently
installed since the former model created a lot of problems. There is also a manual screener
available. If the screener is not functioning properly it affects the sedimentation basin in form
of bigger particles. The screener removes bigger pieces in the incoming water by size
exclusion. The particles attach to the screener and are later burnt.
9.1.2 Activated Sludge
The activated sludge process is a biological treatment process, relying on microorganism’s
removing/converting pollutants. The growth and understanding of the microorganism is
therefore in focus for these methods. Biological treatment can function either as an aerobic-
or anaerobic processes and is proven effective in removing nitrogenous and organic matter
(Gray, 2004).
The activated sludge process in Beaufort West is divided in different zones were aerobic and
anaerobic conditions are predominant. Aerobic conditions are created by aerators, which in
principle are rotating arms, diverting oxygen into the water. The activated sludge is constantly
fed by organic matter in the feed water and converts it into biomass, CO , water and minerals
2
(Gray, 2004).
The sludge age and Sludge Volume Index (SVI) is crucial parameters to monitor in the
activated sludge process. Sludge age is the average time in days the suspended solids remain
in the entire system and SVI is an indication of the sludge settle ability in the final clarifier. If
the sludge age is too high commonly “pin flocs” are formed and particles settles faster in the
second clarifier and the effluent tends to be very turbid. If the sludge age is too low a light and
fluffy sludge is formed commonly called “straggler flocs” which can be observed in the
secondary settler. These problems are long lasting, meaning that they often remain for a week
or more depending on the system (Gray, 2004).
9.1.3 Secondary Settling
The secondary settlers are common Dortmund tanks with a retention time of approximately 4
hours. What comes over the weir of the secondary settlers will end up in the sedimentation
basin, see chapter 9.2.2, and is more or less a result of the treatment efficiency of the activated
sludge. Therefore the activated sludge is the core of the wastewater process. Problems with
excessive sludge going to the sedimentation basin have occurred due to too high sludge age.
31 |
Chalmers University of Technology | Sludge ending up in the sedimentation basin means more frequent backwashing of the sand
filters and UF-membranes in the WRP as well as more frequent cleaning of the sedimentation
basin.
9.2 Treatement Barriers – Wastewater Reclamation Plant
The reclamation system used in Beaufort West (Figure 9.2) can roughly be divided into three
parts: pre-treatment, main treatment and post treatment or polishing. Pre-treatment, which is
the pre-chlorination, sedimentation basin, intermediate-chlorination and rapid sand filtration,
is mainly used to relive pressure of the membranes and prevent fouling. Thereby the life-
length of the membranes is extended, which is highly prioritized since replacing membranes
is connected with high costs. The main treatment is the membrane barriers where the majority
of the pathogens and particles will be separated. The post treatment, UV/H O and final
2 2
chlorination, can be regarded as a safety barrier and used to kill of eventually existing
pathogens. Final chlorination is used to prevent eventual microbiological re-growth in the
pipes and as a protective measure towards re-introduced pathogens.
Figure 9.2 Conceptual model over Beaufort West WRP with on-line monitoring points.
9.2.1 Pre-, Intermediate- and Final-chlorination
The treated wastewater is pre-chlorinated between the secondary settler and the sedimentation
basin. The pre-chlorination point has three major functions:
Disinfection of the feed water.
Hinder algae growth in the sedimentation basin.
Facilitate oxidation of iron and manganese in the sedimentation basin, which will
increase the settling potential.
32 |
Chalmers University of Technology | After the sedimentation basin there is an additional chlorination point, intermediate-
chlorination, which is used if additional disinfection is considered necessary. Final-
chlorination is also used, after the UV/H O treatment, to disinfect the treated water, hinder
2 2
microbiological re-growth inside the pipes and as a safety measure towards pathogens
entering the system after leaving the WRP.
Pre-chlorination has so far shown more benefits to the treatment process then intermediate-
chlorination. During winter, when temperature goes down and algae growth is not as
extensive as during summer, the pre-chlorination can quickly be changed into intermediate-
chlorination.
9.2.2 Sedimentation Basin
When constructing the WRP a new sedimentation basin was also built. The total retention
time for the water inside the sedimentation basin is approximately 18 hours and it has several
important functions:
Work as a buffering zone to handle variations in wastewater flow and composition.
Dilute contaminant peaks.
Increase the settling of particles due to the long retention time.
The long retention time may allow algae to grow inside the river. High amounts of nutrient in
the feed water will increase this risk further, which will result in more frequent cleaning and
backwashing becomes necessary. Cleaning is also necessary to remove settled particles and
sludge (Figure 9.3). When cleaning of the sedimentation basin takes place the feed water may
be by-passed to the rapid sand filters. Without the water passing the sedimentation basin the
rapid sand filters will have to be backwashed more frequent.
Figure 9.3 Cleaning of sedimentation basin
33 |
Chalmers University of Technology | 9.2.3 Rapid Sand Filtration
The coagulation and flocculation process is often commonly followed by a filtration step. By
forcing the water to pass through a granular media, e.g. sand or gravel, suspended solids and
particular matter larger than the pore size of the media are removed. There are two major
filtration techniques, rapid sand filtration and slow sand filtration.
The rapid sand filtration technique (Figure 9.4) is used in Beaufort West’s WRP after the
sedimentation basin. The water passing the rapid sand filter is not aimed at removing any
bacteria or viruses, as in the slow sand filtration, why the processed water often requires
additional treatment to use as drinking water. The flow rate is higher compared to slow sand
filtration and since the rapid sand filter can be put into operation directly after backwash it has
a good cleaning capacity in relation to the required installation area (WHO, 2011b).
Figure 9.4 Conceptual model over typical rapid sand filtration
The rapid sand filters in Beaufort West WRP uses backwash pump blowers to backwash the
filter media by the use of air and water, either separately or simultaneously. The blower
technique will according to Swans water treatment both lower the capital and operational
costs. Three separate pumps are installed to feed the two sand filters with water from the
intake sump. Two of the pumps are working simultaneously and one is on standby during
normal conditions. All backwashing are operated on-site by the operators and the backwash
water from the rapid sand filters is diverted and discharged into the irrigation ponds.
9.2.4 Ultra-filtration
The third barrier in the treatment train consists of ultra-filtration (UF). It is an advanced
process but relies on basic separation principles and is today a proven technique. The UF
consists of membranes and treats the water by physicochemical separation techniques that use
differences in permeability to separate contaminants from the water. The predominant
removal mechanism is straining, or size exclusion. UF refers to the pore size used in the
membranes that are about 0,01 microns in diameter which means that smaller particles then
that will pass through the membranes. Typically UF removes smaller colloids, particles,
sediment, algae, protozoa, bacteria and viruses. Today many membranes are also using
chemical processes for removal of contaminants by adding different coatings. The removal
efficiency of targeted impurities for membrane filtration is typically 99.9999% (6 log
10
reduction) or greater. (Drinking Water Engineering, 2009)
34 |
Chalmers University of Technology | UF is commonly operated as a dead-end flow pattern (Figure 9.5) meaning that the flow
direction is perpendicular to the membrane surface. It is a more simple way of operating the
membranes and requires less energy than cross flow operation. The disadvantage is that the
cake, which consists of accumulated particles, grows with time and consequently the flux
decreases and eventually the dead-end filtration process needs to be stopped to clean, or in
extreme conditions replace, the membranes. The flux decline, due to accumulation of
particles, is one of the main reasons why membrane process remains a challenge both
economically and technically when introduced in a large scale (Swartz, C.D., 2009).
Figure 9.5 Cross flow Operation (a) and Dead-End Operation (b).
All membranes need to be washed when the flux decreases, see figure 9.6. This is typically
done with a repeating pattern. Backwashing means that clean water and/or air is pushed
through the membranes from the opposite side then normal operation. Even if backwashing is
done properly a cake will still build up, which is not removed when backwashing, and the
membranes will typically be operated under constant increasing pressure to produce the same
quantities of water. Once reached a specific threshold chemical cleaning is needed. This can
typically be made without disassembling anything and cleaning can be done in place,
shortened CIP – Cleaning In Place (CIP).
Figure 9.6 Wash cycles of membranes.
35 |
Chalmers University of Technology | The backwash system used in Beaufort West is using blowers, which produces low-pressure
aeration combined with a short reverse flow of filtrate to remove the retained solids from the
membrane fiber bundle. The liquid backwash waste is then drained from the unit to a
backwash waste disposal system. This is done every 20-60 minutes depending on the feed
quality and can be set after a certain time period or when the pressure is increased to a certain
threshold. There will also be an automatic maintenance wash after a preset time interval, or
number of backwash cycles, that require chemicals and more time. This will typically be
made once per week. The CIP is done every month and uses a sodium hypochlorite solution
and acid for cleaning. All of the washing is automated in Beaufort West, but may also be
initiated manually. The backwash water is further diverted to an irrigation pond, while all
cleaning requiring chemicals are diverted to a sludge pond.
The membranes used at Beaufort West are low-pressure membranes with hollow fiber
membrane filtration modules operated from the outside to the inside. Membranes can be
designed to be operated in two different ways, either from the outside to the inside or from the
inside to the outside (Figure 9.7). When operated from the outside to the inside it is generally
easier to maintain the membranes since the area where cake is built up is spread over a larger
area and easier to access, which makes cleaning easier.
Figure 9.7 Membrane operation
When using membrane filtration there is always a loss of water that corresponds to the
backwash water and the concentrate. Typically a recovery rate is specified and for the UF
membranes used in Beaufort West there is a recovery rate of 96%.
UF is commonly used as a “police” for the RO membranes. This means that the UF will
remove a lot of compounds and relieving pressure of the RO membranes since they are more
sensitive. RO is fully capable of treating the wastewater as a single barrier, but resulting in
more backwashing and reduced life length of the RO membranes.
9.2.5 Reverse Osmosis
Osmosis is a specific sort of diffusion. Diffusion occurs when molecules from a region with
higher concentration moves to a region with lower concentration. Osmosis occurs when the
molecules are specifically water and the concentration gradient occurs across a semi-
36 |
Chalmers University of Technology | permeable membrane. The semi-permeable membrane allows water to pass through the
membrane but not ions or larger molecules. Osmosis is thermodynamically favorable and will
continue until equilibrium is reached. The processes can however be stopped and reversed by
increasing the pressure on the concentrated side of the membrane. The pressure must at least
be higher than the osmotic pressure, which is around 27 atm for seawater, to reverse the
osmosis. Due to the high pressure required, reverse osmosis is energy intensive. The water
that is treated is called permeate and refused water is called concentrate or brine. Typical
efficiency of targeted impurities is 50-99%, but RO is overall more efficient than UF if
comparing removed compounds and particles. (Drinking water engineering, 2009)
Beaufort West uses RO membranes (Figure 9.8) that are operated in a cross-flow pattern
(Figure 9.5). Cross-flow has a tendency of less frequent fouling and scaling then dead-end
operation, but requires more monitoring and is overall more complex to operate then dead-end
operation.
Figure 9.8 Membrane operation at Beaufort West. Ultra filtration to the left and reverse osmosis to
the right and BAC filters in front.
The majority of the investment when using filtration/RO techniques is the membranes and it
is therefore crucial to maximize the life length of them. To avoid putting too much pressure
on the membranes, and thereby decreasing the life length, there are mainly three concerns.
Fouling
Scaling (mainly RO)
Chemical attack (mainly RO)
37 |
Chalmers University of Technology | Fouling means that the membrane becomes clogged by organic compounds and the flux
decreases. Fouling occurs due to the load of particles, sediments and microbiological content.
Fouling is inevitable and a fundamental part of the membrane process. When the flux has
decreased to a certain level, backwashing or CIP becomes necessary, see chapter 9.2.4. The
operator may choose to operate the membranes under higher pressure to compensate for
fouling, instead of backwashing. This may pose a risk to the membranes since it may shorten
the life length of the membranes.
Scaling means that insoluble minerals (in-organic compounds) accumulate on the membranes.
Scaling mainly becomes a problem when using RO, since salts become concentrated in the
concentrate. For example if an RO membrane has a recovery rate of 50%, the salt content will
be doubled in the concentrate. This means that a higher recovery rate of the membrane
imposes a higher risk of scaling. Adding chemicals to the feed water can prevent precipitation
of salts. Different chemicals are used depending on the composition of the feed water. If the
main concern is calcium carbonate in the feed water acids can be used and if there are
problems with barium, strontium salts, silicates and iron, anti-scalents can be added. Adding
either anti-scalents or acid to the incoming water when using RO is common praxis today and
means that the salts remains in soluble form and can be separated by the membranes9.
In the case with Beaufort West anti-scalents are being added before the RO. The anti-scalent
added at Beaufort West is specifically designed for the feed water used, and the main concern
is oxidation of iron or manganese. The drawback with using anti-scalents is that it can add to
bio-fouling. Therefore it is necessary to have a gentle balance of the dosing anti-scalents. If
necessary to add anti-scalents it may be essential to reduce the recovery of the membranes to
reduce the risk of over-saturation of precipitating salts10.
Oxidizing agents such as chlorine, bromine, hydrogen peroxide, iodine and ozone causes
chemical attack. A chemical attack is indicated by increasing permeate flow, but with a lower
quality since the damaged membranes passes water and also dissolved minerals. If properly
monitored these situations can be avoided before causing permanent damage. While the UF
membranes are cooping well with chlorine, the RO membranes will be destroyed if in contact
with chlorine for more than a few hours.
For RO conductivity is one of the most critical parameters to measure. The conductivity gives
an indication on how efficient the system is and can be used to trigger alarms if permeate
quality goes down.
RO is connected with modest performance regarding quantities gained after treatment,
typically 50% for seawater to 90% for colored groundwater. Beaufort West loses
approximately 20% of the feed water, meaning that 80% is treated and distributed, which is
rather high for RO membranes. If considering the losses for the UF (4%), the total gained
quantities becomes 76,8%. This is theoretically possible to increase to 85% by recycling some
of the backwash water. The concentrate from the WRP in Beaufort West is diverted to the
irrigation channel. In many cases this may be a problem due to the high salt content and
pathogen load in the concentrate and therefore unsuitable for irrigation. The backwash water
from RO is diverted to the maturation pond where also the activated sludge is diverted.
9 Contractor, Professional Engineer Pierre Marais (WWE), 2011-04-15 (Personal communication)
10 Christopher Wright, Beuafort West Mun. Manager: Technical Services 2011-04-19 (Personal communication)
38 |
Chalmers University of Technology | 9.2.6 UV/H O
2 2
The UV/H O treatment barrier is used for disinfection of the water after the reverse osmosis
2 2
treatment. A mercury UV source with a wavelength spectrum between 200 – 400nm is used
to enlighten the water. The most effective result for water treatment is possible between 200-
280nm (IJ pelaar et al., 2007). The UV radiation is inactivating pathogenic microorganisms,
which also means that they cannot replicate, decreasing the risk for infection further. UV-light
also has the capability to reduce organic micro-pollutants by photolysis, but since this effect is
low H O (hydrogen peroxide) is added to combine photolysis with oxidation to degrade
2 2
micro pollutants (IJ pelaar el al., 2007).
UV is added as an extra precaution if something happens to the RO. When the systems
operates under normal conditions, the permeate from the RO is clean.
9.3 Monitoring
Monitoring means that different parameters, usually connected to water quality, is measured.
The monitoring can either be performed on-line, meaning that results can be seen
momentously, or by physical sampling analyzed in a laboratory, which means that there will
be a delay of the result. Monitoring is a crucial part for any water treatment plant for two main
reasons:
To increase the efficiency of the system and its individual parts.
To use the monitoring result as a quality certificate of the produced water and to
communicate the results to the public and thereby gain trust among the customers.
The sampling is typically performed on in-flowing feed water, between barriers and on the
final water. Ingoing water is monitored to know what pathogens, chemicals, salts etc. that
needs to be removed or decreased in the process and to calibrate the treatment barriers. On-
line monitoring can work as an indicator on the efficiency of the barrier and also alarms may
be connected, while sampling of the final water is used to monitor if the final quality is
adequate for the intended purposes. How extensive the monitoring needs to be is depending
on the composition of the ingoing water and intended use of the final product. When using
ground water as the raw water source, it generally needs less treatment compared to surface
water and has lower tendency to vary and therefore also requires less monitoring.
Skepticism against direct water reclamation is high, why communication of treatment
efficiency and final quality to the public is important, see further information in chapter 9.4.
Therefore an extensive monitoring, beside efficient barriers, is necessary. Monitoring is
usually connected to rather high costs, why there may be implications in a reclamation system
like Beaufort West’s, since the cost for monitoring will be carried by the contractor. Due to
the costs connected with monitoring the contractor may not always want as extensive
monitoring as the client. Further, monitoring is even more important in the beginning of a
project when the treatment plant is new and uncertainties about the treatment process and
corresponding performance are larger.
9.3.1 Monitoring Wastewater Treatment Plant (WWTP)
The monitoring of the wastewater is performed by the municipality and follows suggestions
from Green Drop. No additional monitoring has been added after the construction of the
Reclamation Plant.
39 |
Chalmers University of Technology | The laboratory in Beaufort West is not an accredited lab, resulting in that a yearly sample is
sent to an external laboratory for analysis to fulfill Green Drop standards. More general
monitoring is conducted twice a week by the WWTP manager. Phosphates and E-coli analysis
cannot be performed in the existing laboratory and is therefore analyzed at an external
laboratory on a monthly basis. The parameters that are monitored and the frequency are
presented in Table 9.1. The sampling is done individually for the WWTP and the WRP.
Table 9.1 Parameters measured at Beaufort West’s WRP.
Parameters measured at WWTP
Parameter Twice a week Once a month Once a year
pH
COD
Suspended solids
Electrical conductivity
Total dried solids
Nitrogen
Ammonia
Ortho-phosphate
E-coli
Fluoride
9.3.2 Monitoring Wastewater Reclamation Plant (WRP)
The monitoring program for the WRP is still under development why no final monitoring plan
can be presented in this chapter. However, the existing on-line monitoring will be described
and explained as well as the physical sampling that is performed. Further a suggested
monitoring plan from an external consultant will be presented and compared with the present
monitoring plan.
According to the tender document, the produced water at the WRP must fulfill requirements
for SANS: 241 2005, class 1, i.e. the quality of the drinking water must be acceptable for a
lifetime of consumption. The national standard (SANS: 241, 2005) specifies the quality of
produced drinking water in terms of: microbiological, physical, organoleptic and chemical
parameters. Depending on what is measured they are recommended to be monitored either
daily, weekly, monthly, quarterly or on an annually. The compliance for class 1 is evaluated
on an annual basis, where 95% must fulfill the specified requirement (excluding aesthetic
parameters). Due to increased costs connected to sampling and evaluation it is suggested in
the SANS: 241, 2005 that a graded monitoring system should be implemented. That system
takes into consideration the site specific conditions, e.g. raw water quality, population served,
industrial activities and treatments barriers.
Since the municipality and the contractor together decide which parameters that needs to be
monitored, the monitoring program will be a compromise between cost and safety through
increased monitoring. According to the contractor11 the monthly samples will be done both by
the municipality and the contractor, where the contractor is carrying the cost for the sampling
and monitoring process. The existing tender document suggests that the municipality will do
sampling in the end of the month, while the contractor will do his sampling in the middle of
11 Contractor, Professional Engineer Pierre Marais (WWE), 2011-04-15 (Personal communication)
40 |
Chalmers University of Technology | 9.4 Acceptance - Reclaimed Drinking Water
As indicated from Windhoek, reclaiming wastewater into drinking water may be connected
with problems due to skepticism. Not all people feel comfortable with drinking water
originating from sewage, why open information and communication with the inhabitants is
necessary to build up a confidence and trust towards the drinking water produced.
Buying bottled water is increasingly common in Beaufort West, as throughout the world, and
this poses a threat to the communal tap water as well as the environment. The general picture
is that people originating from the middle class and above tends to drink communal tap water
to a less degree than people from lower class. Information regarding the current available
water and drought situation is good and the necessity of saving water is widespread amongst
the inhabitants. A similar information campaign regarding quality of the produced water is
necessary. Especially when introducing a WRP. This responsibility lies mainly at the
contractor, but is also shared with the municipality. “Water days” and “open house” at the
Reclamation Plant, where school classes and minors are the targeted group, has already been
held with good results14. Targeting youngsters is an effective way to reach the major public,
since they have a more open-minded attitude while elders and parents are more susceptible to
information coming from family.
Producing water from wastewater and get acceptance for the product cannot be done on
routine as when using more conventional treatment. Due to the existing doubts towards
drinking reclaimed wastewater, any minor incidents that can be connected to the reclamation
plant may have big impacts over long time on the confidence of the water delivered. And even
more severe consequences may put the entire project into sank. Trust is gained over decades,
while doubts may develop in seconds.
14 Contractor, Professional Engineer Pierre Marais (WWE), 2011-04-15 (Personal communication)
42 |
Chalmers University of Technology | 10 Risk Assessment for Beaufort West Drinking Water
Reclamation System
In this chapter the risk assessment case study performed for the reclamation system in
Beaufort West, South Africa, is presented. The assessed system is a wastewater reclamation
system that uses treated wastewater as raw water source to produce drinking water.
The risk assessment will be carried out according to the widely accepted risk management
frameworks, see chapter 2, developed by the International Electrotechnical Commission (IEC,
1995) and risk management within the water sector developed by TECHNEAU (2007) as well
as WSP (WHO, 2005). Stakeholders that are considered during the process are:
Employees from authorities and the municipality.
Engineers from the contractor, which are also responsible for operation and
maintenance of the Reclamation Plant.
Drinking water consumers in Beaufort West representing cost-bearers/benefit
receivers and risk takers.
Those that may be affected negatively by the reclamation system e.g. farmers using
treated wastewater for irrigation purposes or inhabitants that may oppose to e.g. drink
enhanced wastewater or increased tariffs.
The risk assessment was, apart from above mentioned stakeholders performed by the authors,
a team of South African water experts, employees from the local government and plant
operators. As new information during the work became available, the different steps were
updated in an iterative process as described in WHO’s WSP. Results from the risk assessment
will be presented in a risk matrix with ALARP levels, ranking all identified risks and a
MCDA that will evaluate suggested risk reduction measures. The risk assessment is only
accounting for present conditions, which means that future planned installations or other
changes will not be considered.
10.1 Risk Analysis
The risk analysis includes:
Scope definition, system boundaries and delimitations (Chapter10.1.1)
Hazard identification (Chapter 10.1.2)
Risk estimation (Chapter 10.1.3)
10.1.1 Scope, System Boundaries and Delimitations
The scope of this risk assessment is to:
Identify hazards threatening the reclamation system and consequently the production
of drinking water, both from a water quantity and water quality perspective, within the
system boundaries.
Estimate risk levels connected to the hazards and rank them by their severity through
the use of a risk matrix.
Define risk tolerability criteria and decide which of the risks that is acceptable and
which need to be reduced.
Suggest risk reduction measures, estimate their risk reducing potential and rank the
measures by the use of a MCDA.
43 |
Chalmers University of Technology | Present a suggestion for a risk reducing implementation plan intended to reduce the
total risk connected to the reclamation system.
The system boundary for the risk assessment includes the wastewater treatment plant and
WRP (Figure 10.1). The starting point of the system is defined as the intake of wastewater to
the WWTP, and the end point is defined as the upstream, blending point, where the produced
drinking water from the Reclamation Plant is mixed with water from the conventional
drinking water treatment. For additional information about the wastewater treatment plant
process and the WRP process, see chapter 9.
Figure 10.1 System boundary for the Reclamation system. Quantities are approximated for average
operation.
As a consequence of the defined boundaries, no other sources, such as boreholes and surface
water, will be considered in the risk assessment, i.e. the mitigating effects this might have on
risks, mainly quantity risks, are not considered in the assessment. This decision was mainly
done to facilitate the risk assessment process and to avoid too high complexity in order to
stress problems originating from the reclamation system. In future WSPs it is however crucial
to consider the entire supply system, i.e. a source to tap approach, as suggested from WHO
(2004) and TECHNEAU (2007).
10.1.2 Hazard Identification
To identify hazards within the system boundaries different techniques were used such as
brainstorming, experience from experts and operators, as well as checklists and databases, see
chapter 4.1 for further information.
44 |
Chalmers University of Technology | To facilitate the process and to be able to present more transparent results, the hazards were
divided in four main categories:
Organization hazards - Management of the plants, funding and contract issues.
Source hazards – The WWTP process is considered as the raw water source.
Treatment hazards – The WRP process is considered as the main treatment.
Future hazards – Problems that may result in future risks.
The source and treatment hazards were further divided into subsystems and all identified
hazards were evaluated, by the help of experts, to determine which of them actually posed a
threat to the reclamation system, i.e. were relevant to be evaluated in the risk assessment.
Approximately 70 hazards needed further investigation.
10.1.3 Risk Estimation
The risk estimation was performed in cooperation with South African water experts and
operators. All hazards judged relevant to the reclamation system were assessed from a quality
and quantity risk perspective and the probability of each hazard was estimated with
predefined categories, suitable for this risk assessment case study, defined by the WHO
(2005) (Table 10.1). Definitions used to estimate quality consequences were also taken from
WHO (2005) and, to include quantity related consequences in the assessment as suggested
from TECHNEAU (2007), similar definitions where developed by the authors (Table 10.2)
All consequence estimations performed in this case study are made from following
perspectives:
Water quality consequences refer to threats towards human health derived from the
consumption of drinking water, i.e. mortality, morbidity or aesthetic impacts.
Water quantity consequences are defined as process downtime of the WRP, meaning
that no water is delivered to the consumers within a certain time interval.
Table 10.1 Definitions of probability categories (WHO, 2005).
Probability
Level Descriptor Description
1 Rare Once every 5 year or has never occurred
2 Unlikely Once per year
3 Moderately likely Once per month
4 Likely Once per week
5 Almost certain Once a day
Furthermore, since the sedimentation basin has a retention time of more than 18 hours, it
functions as a quantity buffer for the WRP. To consider this in the risk estimation a
simplification was made that a quantity failure in the WWTP process must be longer than 24h
before it affects the WRP process, i.e. if the WWTP process is down for 23h it will not have
any effect on the reclamation process and will therefore not be considered as a quantity risk.
The simplification that disruptions in the WWTP process shorter than 24h not will affect the
WRP process implies that these events will not be presented in the risk matrix. The result
from the risk estimation is visible in Figure 10.2 and Figure 10.3 where risks connected to the
WRP has the label “R” and risks connected to the WWTP the label “W”. All risk that was
45 |
Chalmers University of Technology | evaluated as acceptable in the risk tolerability decision (Chapter 10.2.1) has been left out from
the risk matrices.
Table 10.2 Definitions of the quality and quantity consequences connected to each assessed hazard.
Consequence
Level/Descriptor Quantity Quality (WHO, 2005)
1 Insignificant No detectable impact No detectable impact.
2 Minor 0.5h-3h process Minor aesthetic impact causing
downtime dissatisfaction but not likely to lead to use of
alternative less safe sources.
3 Moderately likely 3h-24h process Major aesthetic impact possibly resulting in
downtime use of alternative but unsafe water sources.
4 Major 24h-7 days process Morbidity expected from consuming water.
downtime
5 Catastrophic More than 1 week Mortality expected from consuming water.
downtime
10.2 Risk Evaluation
The result from the risk analysis, i.e. the quality and quantity risk matrices, where used as the
base for the risk evaluation. The risk evaluation includes:
A risk tolerability decision (Chapter 10.2.1), performed with the ALARP approach.
Quantification of risk levels (Chapter 10.2.2), performed using risk priority numbers.
Suggestions and analysis of risk reduction measures (Chapter 10.2.3), performed with
a Multi Criteria Decision Analysis (MCDA).
10.2.1 Risk Tolerability Decision
To increase the understanding of quantity related consequences a separate risk matrix was
created, expressing the risk level (R) in Customer Minutes Lost (CML), see chapter 4.3. Since
the WRP is defined as the only raw water source (Chapter 10.1.1) the proportion of the
population affected, C , will always be equal to 1. This means that the entire population is
A
considered affected if the WRP is not producing water. CML is then expressed as:
R (CML) = P ·1 [3]
F
P corresponds to the probability of failure, used to calculate estimated minutes of production
F
standstill per year. By evaluating the result from the CML matrix (Table 10.3) it is possible to
see that risks occurring more frequent (probability), with shorter downtime (consequence),
may have a longer total downtime per year, then risks occurring less frequent with a more
consistent downtime. Due to the defined system boundaries (Chapter 10.1.1), reservoirs are
not considered in this case study. If the entire water supply system for Beaufort West had
been evaluated, shorter downtimes had been compensated by the reservoir and consequently
would consumers not be affected by shorter downtimes. Therefore a longer continuous
downtime will be evaluated as more severe than a short downtime that occurs more frequent,
even if the CML value is equal.
46 |
Chalmers University of Technology | Table 10.3 The quantity risk expressed as the average CML per year (Customer Minutes Lost), i.e. a
quantitative measure of process downtime. The estimated probability (P) and the consequence (C)
(average process downtime per incident) are used to calculate P . (CML = P *C )
F F A
Insignificant Minor Moderate Major Catastrophic
Almost certain
5,475 27,375 229,950 All year All year
Likely
780 3,900 32,760 224,640 All year
Moderately
likely 180 900 7,560 51,840 120,960
Unlikely
15 75 630 4,320 10,080
Rare
3 15 126 864 2,016
By the use of the ALARP method the risk matrices were (Figure 10.2 and Figure 10.3)
divided in three different risk levels (acceptable risk, ALARP region and unacceptable risk).
The division between the three different risk levels was mainly decided through discussion
among the authors, but opinions from experts in the water sector were also considered15. The
technique used to decide the risk levels was to assess and evaluate the severity of all possible
risk combinations from Table 10.1 and Table 10.2. Questions asked was e.g. if a quality
consequence resulting in mortality occurring once every fifth year acceptable or not?
Since quality and quantity consequences were defined differently (Table 10.2), individual
ALARP levels were chosen for each consequence type. By considering all possible risk
combinations and the result from the CML matrix, risk levels were implemented together with
the risk matrices (Figure 10.2 and Figure 10.3). As can be seen in the figures, only yellow and
red risks are presented, this is due to that risk in the green area not need to be considered
according to the ALARP method. Commonly risks with a more definite effect on health are
looked worse upon then indirect risks with a more vague definition. Drinking water produced
from wastewater is for many connected with skepticism and changes in odour and taste will
be directly linked to the reclamation process, even if the reason behind lies elsewhere. The
already widespread skepticism towards this type of water system means that even the slightest
quality problem may have devastating effects and big impacts on the confidence for the water
system among the public. This result in that the ALARP region in the quality matrix is larger
compared to the quantity matrix. Consequently, the green risk region in the quantity matrix is
larger, indicating that a quantity problem is evaluated less severe compared to if low quality
water reaches the consumer. This means that if there is a minor quality problem in the process
it is advised to shut down the plant instead of delivering low quality water.
The reason that the red risk region was chosen equally large in the two matrices is, despite the
previous discussion, that a severe quantity problem will damage the thrust among the
consumers for the reclaimed water in the same way as problem connected to bad quality,
resulting in that the consumers may look for an alternative drinking water source. If people
lose confidence of the delivered water and stop drinking the water, either for quantity or
quality issues, the whole concept with water reclamation may be jeopardized.
15 Professional Engineer Chris Swartz, Water Utilization Engineers and Ass. Prof. Thomas Pettersson, Chalmers
University of Technology. (Personal communication)
47 |
Chalmers University of Technology | Insignificant Minor Moderate Major Catastrophic
Almost certain R10
Likely
Moderately
likely
Unlikely R1
R3, R4, R5, R6,
Rare W1, R2
R7, R8, R9
Figure 10.2 ALARP regions for quality related risks implemented in a risk matrix. Low risks in the
green region have been left out. R is risks associated with the WRP and W are risks associated with
the WWTP.
Insignificant Minor Moderate Major Catastrophic
Almost certain R26 R10
Likely R25
W10, W11,
Moderately
W12, R21, R22, W3, R11
likely
R23, R24
W8, W9, R16,
Unlikely R1, R30 R17, R18, R19, W2
R20
W1, W4, W5,
W13, W14, R4,
W6, W7, R2,
Rare R6, R27, R28,
R3, R7, R12,
R29
R13, R14, R15
Figure 10.3 ALARP regions for quantity related risks implemented in a risk matrix. Low risks in the
green region have been left out. R is risks associated with the WRP and W are risks associated with
the WWTP.
The risk matrices presents the severity of each identified hazards, according to the defined
ALARP levels. According to the defined ALARP levels, one quality risk and four quantity
risks were evaluated as high risks, i.e. unacceptable and needs to be reduced without
exception. The yellow ALARP region includes risks that need to be reduced if economically
and technically reasonably, and consist of 10 quality risks and 37 quantity risks. Due to
48 |
Chalmers University of Technology | At present the plant is producing 1 Ml/day, which according to the contractor18 can be
produced in approximately 8h, implying that the process is not that vulnerable to staff
absenteeism, i.e. that it will be enough with one operator working an 8h shift to produce the
required amount. However, since the plant will increase the produced quantity with 10% a
year for the next 10 years (according to the agreement between the contractor and the
municipality), the plant will have to run for approximately 20h per day to produce the
required capacity in year 2020. Consequently the plant needs more operators in the future to
produce the required quantities. The possibility to handle less likely circumstances, as deaths
or operators changing work is also small with the present working force.
10.2.2 Risk Priority Number
After the ALARP levels were set, the scales of the consequence and probability axes were
defined. The reason that this was not done in the opposite order is because the ALARP levels
had been defined trough reasoning and discussion. The numbering of the scales will just make
it possible to present the risk in quantitative terms, i.e. with the risk priority number (Chapter
4.2.).
The Water Research Commission (WRC) recommends a nonlinear inclination of the
probability and consequence scales (Figure 10.6). From the figure it can be seen that the
consequence and probability curves has the steepest inclination in the middle span. This
means, that a reduction of the consequence and/or probability in that span have the largest
effect on the risk priority number. Consequently, the result of defining the scales according to
the WRC suggestion will be that risk reduction measures are focused towards the medium
probabilities and consequences. To avoid a higher focus on less severe risks, an exponential
scale (Figure 10.6) is proposed and implemented in this case study. The use of an exponential
scale will benefit risk reduction of high risks.
Figure 10.6 Non-linear inclinations of the probability and consequence scales as suggested by the
WRC (left). Exponential inclination of the scales as used in this case study (right).
18 Professional Engineer Pierre Marais, Contractor, (WWE), 2011-05-19 (Personal communication)
52 |
Chalmers University of Technology | The scales and risk priority number are presented together with the ALARP levels in Figure
10.7 and Figure 10.8. As can be seen the risk priority number is not uniformly distributed in
the risk matrices. A result of that a decrease in consequence is considered more favorable than
a decrease in probability considering the total risk reduction. By applying equation [1]
(Chapter 4.2) the consequence scale has been weighted higher (b=1.6) than the probability
scale (a=1.0).
Risk Profile
Low (0 < X ≤ 8) Acceptable risk
Medium (8 < X ≤ 84) The risk can be acceptable if it is economically and technically
unreasonable to reduce it.
High (X > 84) The risk cannot be accepted under any circumstances.
Consequence
1 2 4 8 16
16 16 49 147 446 1351
y
t 8 8 24 74 223 676
ilib
a 4 4 12 37 111 338
b
o
r 2 2 6 18 56 169
P
1 1 3 9 28 84
Figure 10.7 ALARP regions and risk priority number for water quality related risks. R=P1.0 * C1.6.
Risk Profile
Low (0 < X ≤ 12) Acceptable risk
Medium (12 < X ≤ 84) The risk can be acceptable if it is economically and technically
unreasonable to reduce it.
High (X > 84) The risk cannot be accepted under any circumstances.
Consequence
1 2 4 8 16
16 16 49 147 446 1351
y
t 8 8 24 74 223 676
ilib
a 4 4 12 37 111 338
b
o
r 2 2 6 18 56 169
P
1 1 3 9 28 84
Figure 10.8 ALARP regions and risk priority number for water quantity related risks. R=P1.0 * C1.6.
10.2.3 Multi Criteria Decision Analysis (MCDA)
In the risk evaluation process (Chapter 10.2) five unacceptable risks, connected to the
reclamation, system were identified (Table 10.4). Since unacceptable risks, according to the
ALARP-method, cannot be accepted under any circumstances, several risk reduction
measures were suggested by the authors (Table 10.5). It is of importance to consider that a
risk reduction measure may affect other additional risks than the target risk, e.g. measure R1.1
53 |
Chalmers University of Technology | that also affects risk R3, R10 and W11, which needs to be considered in the MCDA to attain a
more complete result.
Table 10.4 Identified unacceptable risks
Risk Description Risk type
W2 Intake screws in the WWTP become damaged or stops functioning, Quantity
leading to no flow and consequently no feed water to the WRP
W3 Inadequate floc settling in the WWTPs secondary settler. Quantity
R1 Inadequate monitoring resulting in water quality or quantity risks. Quality/Quantity
R10 Inadequate local knowledge of operation and condition of the Quantity/Quality
installation.
R11 Not sufficient numbers of staff/operators. Quantity
Table 10.5 Risk reduction measures suggested by the authors.
Target Additional events
Ref Risk reduction measure desciption risk affected
W2.1 Install a second screw pump W2 -
W3.1 Install an additional secondary settler W3 (W12)
W3.2 Increased monitoring of sludge age and FeCl dosing. W3 (W10, W13)
W3.3 Replace flow control of recycle pumps to more reliable W3 (W9)
technique.
W3.4 Combination of measure W3.2 and W3.3 W3 (W9, W10)
R1.1 Installation of adequate laboratory + increased monitoring R1 (R3, R10, W11)
program (according to suggestions by C.D. Swartz) and
education in sampling of the operators.
R1.2 Increased monitoring program (according to suggestions R1 (R3, R10, W11)
by C.D. Swartz) and education in sampling of the
operators, samples sent away for analysis.
R10.1 Educate all operators to a minimum level to run the plant R10 (R11)
R10.2 Increase the knowledge of the operators to a higher level. R10 (R3, R11, R20)
R10.3 Secure the possibility to get support from external source R10 -
R10.4 Develop operation manuals for the treatment process R10 -
R10.5 Combination of measure R10.1 and R10.4 R10 (R11)
R11.1 Employ one more operator R11 -
R11.2 Employ two more operators R11 -
R11.3 Increase salaries and better working conditions. R11 -
R11.4 Combination of measure R11.1 and R11.3 R11 -
R11.5 Combination of measure R11.2 and R11.3 R11 -
Further, all risk reduction measures have different potential to achieve an acceptable risk level
for its target risk, i.e. that the final risk level after risk reduction is in the acceptable/green area
of the risk matrix. The probability to achieve an acceptable risk level is larger for a measure
that reduces the target risk into the green area compared to a measure only reducing the same
target risk into the yellow area of the risk matrix. Since there also are uncertainties involved in
54 |
Chalmers University of Technology | the estimation of risk reduction, a MCDA method developed by Lindhe (2010) was used in
the case study. The MCDA method uses beta distributions (Table 10.6) to model uncertainties
connected to the estimation of a measure’s probability to achieve an acceptable risk level. α
and β are numerical parameters used to define the shape of the beta distributions. To combine
the involved uncertainties with the probabilities of different measures to achieve an
acceptable risk level, the quality and quantity risk matrices were divided in different
probability categories (Figure 10.9). The categories reflect the probability that a measure is
achieving an acceptable risk level for its target risk, i.e. the lower the final risk level in the
risk matrix is for the target risk, the higher is the probability that the risk is reduced to an
acceptable level.
Table 10.6 Probabilities to achieve an acceptable risk level modelled by beta distributions depending
on final risk level position in risk matrix. α + β = 42, meaning that uncertainties for all distributions
are equal. (Figure 10.9) (Modified from Lindhe, 2010).
Risk priority number (category Probability of achieving an
from Figure 10.9.) acceptable risk level α β
Quantity Quality Most likely P05 P95
224-1351 (VII) 224-1351 (VII) 0.00 0.00 0.07 1 41
85-223 (VI) 85-223 (VI) 0.10 0.05 0.21 5 37
38-84 (V) 38-84 (V) 0.30 0.20 0.43 13 29
13-37 (IV) 9-37 (IV) 0.50 0.37 0.63 21 21
7-12 (III) 5-8 (III) 0.70 0.57 0.80 29 13
3-6 (II) 2-4 (II) 0.90 0.79 0.95 37 5
1-2 (I) 1 (I) 1.00 0.93 1.00 41 1
a) Quality Insignificant Minor Moderate Major Catastrophic
IV V VI VII
Almost certain
III V VII
Likely
IV VI
Moderately likely
II
III IV V VI
Unlikely
I II IV V
Rare
b) Quantity Insignificant Minor Moderate Major Catastrophic
IV V VI VII
Almost certain
III IV V VII
Likely
VI
Moderately likely II III
IV
V VI
Unlikely
I II
III IV V
Rare
Figure 10.9 Categories defining the probability (Table 10.6) to achieve an acceptable risk level for the
target risk, depending on the position in the risk matrix. I = Highest probability, VII = Lowest
probability. a=Quality related risks, b=Quantity related risks.
55 |
Chalmers University of Technology | Further the MCDA also considers uncertainties involved in the estimation of cost to
implementing a measure, which was modelled by discrete distributions (Table 10.7).
Normalized values between 0-1 were used in the estimation, where 0 represent the highest
cost and 1 the lowest cost. Consequently favouring measure with lower cost when calculating
the final performance score. Table 10.8 presents the estimated risk reduction of the target
risks and Table 10.9 presents the estimated risk reduction of the additional risks, both used in
the MCDA. When calculating the total benefit, risk reduction in water quality and quantity is
weighted as equally important. It is of importance to remember that uncertainties (Table 10.6)
only are estimated for the risk reduction of the target risk, but assigned equally if additional
risks are affected by the measure.
Table 10.7 Discrete distributions used to model uncertainties in cost estimations (Lindhe, 2011).
Probability of each cost category
Cost Low Low/medium Medium Medium/high High
Low 0.68 0.16 0.09 0.04 0.03
Low/medium 0.10 0.70 0.10 0.06 0.04
Medium 0.05 0.10 0.70 0.10 0.05
Medium/high 0.04 0.06 0.10 0.70 0.10
High 0.03 0.04 0.09 0.16 0.68
Table 10.8 Risk reduction/benefit for target risks expressed in the risk priority number.
Risk priority number
Quantity Quality Risk reduction/benefit
Ref
Initial After Initial After of target risks
W2.1 169 28 - - 141
W3.1 111 9 - - 102
W3.2 111 56 - - 55
W3.3 111 56 - - 55
W3.4 111 9 - - 102
R1.1 18 9 169 3 175
R1.2 18 9 169 3 175
R10.1 147 74 16 4 85
R10.2 147 12 16 1 150
R10.3 147 147 16 16 0
R10.4 147 74 16 16 73
R10.5 147 37 16 4 122
R11.1 111 18 - - 93
R11.2 111 9 - - 102
R11.3 111 111 - - 0
R11.4 111 9 - - 102
R11.5 111 9 - - 102
56 |
Chalmers University of Technology | Table 10.9 Risk reduction/benefit for additional event affected expressed in the risk priority number.
Risk priority number
Add. event Quantity Quality Risk reduction/benefit
Ref
effected Initial After Initial After of add. risks
W3.1 W12 37 12 - - 25
W3.2 W10 37 2 - - 35
W3.2 W13 28 9 - - 19
W3.3 W9 56 28 - - 28
W3.4 W9 56 28 - - 28
W3.4 W10 37 2 - - 35
R1.1 R3 28 28 28 3 25
R1.1 R10 147 74 16 8 81
R1.1 W11 37 6 - - 31
R1.2 R3 28 28 28 3 25
R1.2 R10 147 74 16 8 81
R1.2 W11 37 6 - - 31
R10.1 R11 111 56 - - 55
R10.2 R3 84 28 28 3 81
R10.2 R11 111 56 - - 55
R10.2 R20 56 28 - - 28
R10.5 R11 111 56 - - 55
A spreadsheet was constructed to perform the calculation for the MCDA and the result is
presented in Table 10.10. When calculating the performance score, see equation [5], Monte
Carlo simulations were used to handle the beta distributions. Uncertainties were assigned to
the estimation of the benefit, see equation [4], and to the cost of implementation. Furthermore,
the risk reduction and cost were weighted as equally important.
Benefit = (B + B ) * B [4]
T A D
Score= 0.5*Benefit (normalised) + 0.5* Cost [5]
B = Benefit of target risk (Table 10.8)
T
B = Benefit of add. risks (Table 10.9)
A
B = Beta distribution (Table 10.6)
D
Score = Performance score (Table 10.10)
Cost = Cost of implementation (Table 10.7)
The last column of Table 10.10 describes the probability that the final risk level is higher than
the acceptable risk level and is used when choosing between measures with similar risk
reduction and cost.
57 |
Chalmers University of Technology | Table 10.10 Multi Criteria Decision Analysis (MCDA) performance matrix, using beta distributions.
Benefit= risk reduction/benefits, Score= Performance score, R= Initial risk level, R = Final risk
I F
level, R = Critical risk level, P(R > R ) = Probability that the final risk level is higher than the
C C
critical/acceptable risk level.
R -> R P(R > R )
I F C
Measure Benefit Cost Score qual. quant. qual. quant.
W2.1 71 Low/Medium 0.48 - - 0.50
W3.1 88 High 0.25 - - 0.31
W3.2 34 Low 0.49 - - 0.69
W3.3 26 Low 0.48 - - 0.69
W3.4 114 Low/Medium 0.57 - - 0.31
R1.1 253 Medium/High 0.66 0.12 0.31
R1.2 253 High 0.57 0.12 0.31
R10.1 51 Low 0.53 0.12 0.69
R10.2 229 Low/Medium 0.80 0.02 0.31
R10.3 0 Low 0.43 - -
R10.4 23 Low 0.47 - 0.69
R10.5 94 Low 0.61 0.12 0.50
R11.1 47 Medium 0.34 - - 0.50
R11.2 71 Medium/high 0.29 - - 0.31
R11.3 0 Low/Medium 0.35 - - -
R11.4 71 Medium 0.39 - - 0.31
R11.5 71 High 0.21 - - 0.31
10.3 Chain of Events
When using a multi-barrier concept, risks that require chain of events to occur, meaning that
failing components or events interacts, are more frequent since there are several barriers
aimed at treating the same targeted impurities. In a reclamation system, including a WWTP
and the WRP itself, the interaction between the WWTP and the WRP is important. Many
hazards that can be connected to the WWTP require a chain of events, proceeding into the
WRP, to become a valid risk to the system. To illustrate this there are good tools, e.g. Fault
tree and Event tree, but these typically requires more data and they are more of a detailed
study. Wastewater also has a tendency to vary, both over short periods as well as for longer
periods, which also may impact the risk.
10.4 Sensitivity Analysis and Uncertainties
Due to lack of data, the authors, in cooperation with South African water experts, performed
all estimations of probability and consequence. Consequently, uncertainty connected to risk
level estimations need to be considered. The probability is as mentioned before (Table 10.1)
defined trough frequency categories with sometimes large differences between each step, e.g.
moderately likely was defined as “once a month” and the next step, unlikely was defined as
“once a year”. This creates difficulties when the most likely frequency for a hazard occurs
between two categories. One solution to this problem is to always choose the most likely
category, resulting in that probability sometimes is underestimated and vice versa. Another
solution is to always estimate the more frequent of the two categories and thereby
overestimate the risk level. An underestimation of the risk levels is, of obvious reasons, not
appropriate when performing a risk analysis but an overestimation is neither always the best
58 |
Chalmers University of Technology | solution. An underestimation may result in that a risk is overlooked and an overestimation
will result in increased risk reduction cost for some time unnecessary measures.
The solution in this case study was to rather see the probability and consequence as a
combination that together define the risk level (Equation [1], chapter 4.2) and by defining the
hazards more accurate. If a hazard is defined as: “Power failure affecting the WRP”, it is very
hard to estimate the consequence and probability, since a long power failure will have a more
severe effect but a less probability and vice versa. Consequently by defining the hazard as
“Power failure longer then 24h affecting the WRP” the probability and consequence can be
more accurately estimated. When the approach was unsuitable, the more frequent probability
or more severe consequence was estimated to avoid underestimations.
The MCDA model that was used in this case study considers only uncertainties connected to
the estimation of risk reduction of the target risk and the cost of implementing the measure.
Since there also are uncertainties involved in the estimation of the initial risk level (Figure
10.2 and Figure 10.3), a MCDA method that considers this would be useful, e.g. the discrete
MCDA model developed by Lindhe et al. (2010). The sensitivity in the result of the MCDA
was also affected by the assigned importance between the risk reduction/benefit and cost
when the performance score was calculated. If the cost e.g. was chosen as twice as important
as the benefit, possible in a poor town like Beaufort West, it would be more beneficial to
choose measure R10.1 over R10.2. This implies that the decision-makers must consider what
is most valuable, to reduce the risk as much as possible or to as low cost as possible. A change
of the estimated cost or risk reduction for a measure will also have large impact for the
MCDA result.
10.5 Results from the Risk Assessment
The risk assessment process identified five risks as unacceptable (Table 10.4), additional 47
were identified inside the ALARP region and 29 as acceptable (not presented in the case
study). It is of great importance that the 47 ALARP risks are evaluated further in a future risk
assessment since they might be possible to reduce to an acceptable cost.
The full MCDA result is presented in Table 10.10 and the measures that were evaluated to be
most beneficial, from the defined criteria, is presented in table 10.11. All measures illustrated
in table 10.11 are the ones that have the highest performance score and they will also reduce
the target risk to an acceptable level. However, it is still important that the MCDA result is
analyzed with common sense and trough discussion among the decision makers.
59 |
Chalmers University of Technology | 11 Discussion
It is crucial when performing a comprehensive risk assessment to involve all major
stakeholders, e.g. municipalities and contractors, to a large extent in the process. This will
increase the understanding of the risk assessment concept and facilitate the process to reach
an agreement concerning e.g. ALARP regions and risk estimations. The result from the risk
assessment is not intended to be a surprise for the stakeholders; the process should instead be
as transparent as possible.
Few things can be done on routine. ALARP regions, quantity/quality definitions, scaling of
axes etc. need to be considered at each new risk assessment. Comprehensive risk assessments
are time consuming and rather demanding to perform to all involved participants. If
continuing with more sophisticated analysis as Fault Tree Analysis (FTA) or Event Tree
Analysis (ETA) and uncertainty analysis a higher level of expertise and statistical modeling is
required, further stressing the importance of an active role by the stakeholders. All parts need
to take an active role and the ALARP levels should ideally be decided among involved
stakeholders to avoid conflicts about the results.
WRP’s differ from other conventional water treatment plants mainly due to the origin of its
raw water. Since the raw water originates from treated wastewater, all processes in the
WWTP need to be considered in the risk assessment. If the risk assessment only is focused on
the WRP, risks may be overlooked or underestimated. It is also common that the WRP and
WWTP are operated by different contractors or like in Windhoek divided between the
municipality and a contractor, which may result in conflicts if not a well-defined contract, as
in Windhoek19, specifying responsibilities and commitment is available. The city of
Windhoek20 stresses that a central part for a successful WRP project and future cooperation
has been to have a good contract to rely on. The risk assessment performed in this project has
not analyzed any possible contract issues since the contract concerning Beaufort West WRP
still is under construction. This is a big drawback, since we expected some risks to be related
to contractual issues. In future risk assessments project where different parts/organizations
operates the WWTP and WRP, it is crucial to include the contract in the assessment, it may
also be suitable to handle it in a separate stage.
WRP’s are big investments and rather complicated to operate and the acceptance for the
produced water may initially be low. Further they will never be able to fully replace
conventional water sources since the recovery rate is low considering the whole system,
meaning that losses in the system is large. Therefore WRP’s should not be considered as the
first solution to handle severe water stress. First a thorough investigation of existing sources
and consumption patterns should be performed to see if more simple and less expensive
measures can be taken to either reduce consumption, or if there are other water sources to
explore. Beaufort West did consider all other options and due to acute water stress the
decision of a construction of a WRP came up rather fast. Overall the whole project did not
take more than two years to accomplish. The risk assessment was initiated when the plant was
already producing water. Ideally the risk assessment should have been done before the design
phase to facilitate necessary changes and also relevant information would have been available
earlier in the project.
19 J.G. Menge, Analysis Responsible for Gammams WWTP 2011-03-29 (Personal communication)
20 J.G. Menge, Analysis Responsible for Gammams WWTP 2011-03-29 (Personal communication)
61 |
Chalmers University of Technology | Windhoek that has produced drinking water from wastewater for over 50 years has a lot of
experience in the field. In future reclamation projects in South Africa, as well as the rest of the
world, valuable information from Windhoek should be considered. Also, Beaufort West will
now have a lot of valuable experience for future WRP projects.
Due to the initial rather limited knowledge among the authors concerning the treatment
processes in the WRP, it was tempting to put a lot of focus and confidence on the hazard
databases (THDB and WRC checklist). Hazard databases have an important role to play in
risk assessments to facilitate hazard identification and increase the number of identified
hazards. However, it is important to stress that the databases are developed as a complement
for the hazard identification, only containing the most general hazards. If too large focus is
put on the databases there is a risk that hazards more specific for the assessed system is
overlooked. The hazards in the databases are often also not very well defined, resulting in that
it becomes difficult to estimate accurate consequences and probabilities.
Reverse osmosis (RO) is known as a very efficient treatment barrier, with only few
compounds able to pass through. Therefore quality risks when using reverse osmosis are not
many as long as the RO is functioning as intended, the risk for breakthrough is small since
there exists online monitoring of e.g. pressure. In Beaufort West, extensive alarm systems are
used when the processes is operating under design conditions. When using such alarm
systems there is a potential danger that it trigger too often, resulting in a longer downtime of
the WPR. The RO is the main reason why such a few quality related risks have been found,
but still the RO is a rather sensitive system and it is crucial to ensure adequate pre-treatment
to maximize the life-length of the membranes since they correspond to such a big part of the
investment. Since the UF precedes the RO and thereby needs to take the “first hit”, higher
risks towards the UF-membranes can be expected.
WRP’s using RO needs to deal with rejected water, so called brine, and backwash water. CIP
(Cleaning In Place) often contains chemicals that may be harmful for the environment if
released untreated and the rejected water contains a high pathogen load and/or salt content. In
Beaufort West is the backwash water, that contains chemicals, diverted and collected in
activated sludge ponds and the less toxic backwash water is together with the rejected water
diverted to an irrigation channel. The consequences these might have are not fully covered
and should be further analyzed in a Environmental Impact Assessment.
Many risks were assessed to have a “Rare” probability, with “Catastrophic” consequences,
resulting in that they were to be found inside the ALARP-region (Figure 10.2 and Figure
10.3). These risks are hard to handle since there are limited measures available to reduce the
risk. The only possible parameter to decrease is the consequence, but for some risks, e.g.
earthquakes and flooding, it is extremely uneconomical to do so. Still it makes sense to
illustrate them in the risk matrix to be aware of them, but for some it may look worse than it
is.
As suggested from the case study in Windhoek21, the water from the reclamation system
should be of at least as good or preferably of higher quality than existing sources to increase
acceptance. This makes sense, but when considering the entire water system in Beaufort
West, the water from the reclamation system is blended with water from conventional
treatment, resulting in that water reaching the consumer will have almost the same taste and
21 John Esterhuizen, General Manager: WINGOC 2011-03-29 (Personal communication)
62 |
Chalmers University of Technology | odor as before. In Beaufort West there is a big difference in taste between water from the
conventional system and water from the WRP. The water from the WRP tastes preferably
better than the water from the conventional system. But since the water is mixed in the
reservoirs measures focused on increasing “Minor” consequences (aesthetic impacts) will not
be very beneficial as long as the conventional treatment is not upgraded.
When estimating quality consequences it is hard to distinguish between what incidents that
may result in aesthetic impacts, and which might lead to illness, death etc. This since it then
must be connected to what pathogens that may pass through the system. Ones again this is
often requiring a chain of events. In the risk assessment quality related risks have been
assessed from how severe impacts to the process the hazard may result in and for how long
time may insufficient water be delivered to customers instead of a deeper analyzes including
what different pathogens that may pass the system under certain conditions.
Multi Criteria Decision Analysis (MCDA) is a systematic approach to rank risk reduction
measures, and provides a good background towards choosing the most appropriate measure(s)
from a decided set of criteria. Even though, an MCDA does not give a final result, but more
decision basis. The final decision on what risks to prioritize, and which risk reduction
measures to choose, will still be a combination of sound sense and by remarking the results
from the risk matrices and the MCDA analysis.
63 |
Chalmers University of Technology | 12 Conclusions
Conclusions in this chapter are divided in two parts. One part that concerns the risk
assessment case study performed for Beaufort West’s wastewater reclamation system and a
second part were general conclusions concerning wastewater reclamation is presented.
Conclusions concerning the Beaufort West WRP risk assessment case study:
Monitoring is crucial for any reclamation system. Beaufort West’s WRP is not having
adequate, monitoring due to lack of funding. An expansive monitoring plan needs to
be integrated in the initial budget plan to avoid this situation. Monitoring should not
come in second hand
A second screw pump needs to be installed to decrease the probability that problems
occur in connection to the WWTP intake.
The sludge age in the secondary settler and the FeCl dosing needs to be monitored
3
more frequent. This in combination with the replacement of the recycle pumps flow
controls to a more reliable technique will lower the probability for inadequate settling
in the secondary settler.
The knowledge of the operators needs to be increases to a higher level to minimize the
probability that the WRP process stands still due to operator error.
At least one more operator needs to be employed to decrease the vulnerability for
sickness, unaccepted deaths etc.
An ISO-certification for Beaufort West’s WRP should be initiated to minimize risks
for human errors and external parts may stress weaknesses in the operative system.
Due to limitations of the risk matrix method handling chain of events, further studies
that illustrates the interaction between the WWTP and the WRP is highly appropriate.
By doing so, some risks may be found not applicable, while some new risks may
appear.
Sufficient monitoring to fulfill Green Drop is already established for the WWTP. To
increase efficiency, and decrease costs, of monitoring for the contractor of the WRP,
there are opportunities to incorporate and share some parts of the monitoring with the
WWTP.
Due to delimitations, risks in this report have been focused towards stressing the
reclamation system. In future Water Safety Plans (WSP) the whole system, including
the conventional WTP, groundwater sources, distribution system and reservoirs needs
to be analyzed from a quantity perspective.
Overall conclusions regarding wastewater reclamation:
An expertise group, gathering knowledge concerning direct- and possibly also
indirect, wastewater reclamation systems should be established in southern Africa.
Beaufort West Municipality and the operator of the WRP should set up frequent
meetings with relevant participants from Windhoek’s reclamation system to share
information and experience.
Ideally the WWTP and the WRP is operated by the same part to fully utilize the
benefits of one system instead of needing to handle the WWTP and WRP as separate
systems.
A well-defined contract, specifying which part that is responsible for what, should be
prepared before construction.
64 |
Chalmers University of Technology | Reverse osmosis is a very efficient treatment barrier able to treat almost all types of
water. However there is a potential danger of having too high confidence on reverse
osmosis. To ensure adequate performance connected alarms to the different barriers is
an efficient way to increase the life-length of the membranes and to ensure an
adequate quality of the water.
Using databases when performing risk analysis is an efficient way to incorporate as
many potential hazards as possible. Still it should only be used as a basis for
discussions. The hazards are defined very general in THDB, which has a point, but
when assessing consequences and probability it is crucial to know specifically what
consequence that is addressed to be able to estimate correct probabilities. A good and
simple way to handle this problem is to discuss the consequences, before probabilities.
It is important to incorporate water quantity problems, as suggested by TECHNEAU.
If there are frequent, or consistent, interruptions of the water supply it may result in a
low confidence of the water system and consequently the consumers will look for
other water sources.
65 |
Chalmers University of Technology | Risk
Hazard Hazard description Comments priority
number
Transmembrane pressure (TMP) alarm should pick this
up, also possible to noticed during maintenance.
R6 RO: Membrane break trough; fiber breakage. Possibility to repair smaller breakage on their own at site. 28
Since this is final membrane, quality may be affected if
the alarms do not work.
RO: Reduced filter
performance/destruction/membrane
R7 The membranes are tested every month. 28
imperfections leading to that the water cannot
be cleaned to class 1 quality.
The water is final-chlorinated (maximum allowed level of
Re-contamination of the water in the pipeline
R8 0.8mg/l is the aim) to hinder biological growth and re- 28
before reaching reservoir.
contamination in the pipelines. New built pipeline.
The alarm system is connected to the on-line monitoring.
If the monitoring system fails the plant will shut down,
due to that the PLC requires information from the
monitoring to run the process. Alarms may however
R9 Failure of the alarm system. malfunction, and thereby cause quantity/quality 28
problems due to that the failure is not communicated to
the PLC or operator. Problems with to sensitive alarms,
causing quantity failures, are at the moment adjusted by
the contractor.
The knowledge on site is at the moment not adequate.
There is only one operator that has training and can take
necessary decisions. If the operator not can solve a
problem, people from Cape Town need to come to the
Inadequate local knowledge of operation and site, which is five hours away. Operators have no
R10 16
the condition of the installation. education in sampling and only trust the on-line alarms.
For now the operators calls the responsible installer for
the membranes almost on a daily basis. There is no
backup if the trained operator gets sick, or decides to quit
the job. |
Chalmers University of Technology | Risk
Hazard Hazard description Comments priority
number
A long severe drought can affect the amount of wastewater
Quantity related problems (e.g. water shortage reaching the WWTP and consequently the feed water later
W7 84
leading to closing of intake). reaching the WRP. Salinity becomes high if there is a
drought, problem to RO.
There are backup pumps for all treatment steps (excl screw
pump) and there is local mechanical knowledge in town.
W8 Pump failure leading to downtime. There has been an incident were the activate sludge recycle 56
pumps failed leading to that sludge was carried over to the
sedimentation basin.
Pump with flow control decides how often the sludge should
Improper operation or inadequate desludging
W9 be removed. The flow control has failed resulting in sludge 56
programme.
was carried over in the secondary settler.
Dosing malfunction (FeCl) due to human
3
errors or mechanical failure, can reduce floc Operators are monitoring so the dosing works at least once
formation and thus result in inefficiently a day. Malfunction of FeCl can lead to more backwashing,
W10 3 37
remove of harmful microorganisms, organic CIP etc. The dosing of FeCl is decided from flow and
3
material, color and turbidity (too low/high phosphate concentration in incoming wastewater.
dosing).
Non-optimized treatment in the WWTP will result in poorer
quality of feed water and trigger more alarms in WRP and
Non-optimized treatment processes can result
W11 require more frequent backwashing or CIP. Algae from 37
in poor process performance.
sedimentation basin create biofilm growth on RO
membranes leading to CIP etc.
Spikes during rain and too high sludge age can result in
Large quantities of storm water disturbing
overflow, resulting in that the sedimentation basin need
W12 process, leading to over flow of activated 37
more frequent cleaning as well as more frequent
sludge process.
backwashing.
Daily monitoring of sludge age. Contractor wants a longer
W13 Unappropiate sludge age. 28
sludge age then the municipality.
Thieving and stealing of fencing is a fact. One guard is
The site is not secure leading to theft of
posted at site during night, and needs to cover a large area.
equipment resulting in downtime of more than
W14 Has limited possibilities to see what is happening with the 28
24h (i.e. no fencing, gates, locks,
fencing. Stealing of parts in the WWTP is expected to have
safety/warning signs, inadequate security).
less quantity consequences then stealing at the WRP. |
Chalmers University of Technology | Risk
Hazard Hazard description Comments priority
number
Water Reclamation Plant (WRP)
The knowledge on site is at the moment not adequate.
There is only one operator that has training and can take
necessary decisions. If a problem not can be solved by the
operator’s, people from Cape Town need to come to the
Inadequate local knowledge of operation and
R10 sight, which is five hours away. Operators have no education 147
the condition of the installation.
in sampling, but only rely on the on-line alarms. For now the
operators calls the responsible installer for the membranes
almost on a daily basis. See chapter 10.2.1 for more
information.
Only one operator that is trained enough, and two in total.
Already they work long passes, and there are very limited
possibilities to run the plant if one becomes sick. Especially if
the trained operator becomes sick. The contractor will in
R11 Not sufficient numbers of staff/operators. the future train both operators to an adequate level and 111
educate the "cleaner" to basic level of process control.
There is no backup if the trained operator gets sick, or
decides to quit the job. See chapter 10.2.1 for more
information.
Vandalism or sabotage may pollute the water
with chemicals or microbes or damage
R2 Highly unlikely. Can be decreased by increasing security. 84
equipment and infrastructure, leading to
severe/long process interruption
Operators will call contractor on all problems. All alarms are
Issues of concern are not addressed due to noted in a book. Sand filters are operated manually, no
R3 inadequate reporting (e.g. malfunctions, possibility for online monitoring. PLC interface can be 84
compliance reports). handled from Cape Town as well. Contractor on site once a
month.
Responsible for membranes says that the 5 years guaranty
RO: Reduced filter
will be fulfilled, no problem. TMP alarm will trigger if
performance/destruction/membrane
R7 something is wrong with the membranes. If membranes 84
imperfections leading to that the water cannot
need to be replaced they will have to be delivered from
be cleaned to class 1 quality.
outside the country.
Geophysical accidents (e.g. earthquakes, Lightning damaging electrical equipment, long reparation
R12 84
flooding, landslides, lightning). time. Highly unlikely. Stabile area.
Responsible for membranes says that his 5 years guaranty
UF: Reduced filter will be fulfilled, no problem. Problems if membranes are
performance/destruction/membrane operated/maintained out of spec. TMP alarm will trigger if
R13 84
imperfections leading to that the water cannot something is wrong with the membranes. If membranes
be cleaned to class 1 quality. need to be replaced they will have to be delivered from
outside the country. |
Chalmers University of Technology | Risk
Hazard Hazard description Comments priority
number
R14 Fire (e.g. due to smoking inside the plant). Smoking is taking place inside the plant. 84
Alarm monitoring for residual chlorine before RO. If
RO: Residual chlorine damaging the
R15 membranes need to be replaced they will have to be 84
membranes.
delivered from outside the country.
RO: Failure of the compressor; pneumatic Spare parts are not available on site, knowledge in Cape
R16 56
system or of other installation hardware. Town.
There are backup pumps for all treatment steps. But local
knowledge may not be enough to fix the problem, e.g. when
they had a pipe burst in the inlet pump room and the room
Hydraulic failure (e.g. pipe burst, pipe failure,
R17 was flooded, leading to that 2 pumps were needed to be 56
pump failure) leading to process downtime.
sent to George for repair. Always a risk of pipe burst of
distribution pipeline. Flow meters with connected alarms
will trigger if out of spec.
Problems with frequent thieving in the area and stealing the
fencing are a fact. In the WRP there is more valuable
equipment that is more desirable and also necessary for the
The site is not secure (i.e. no fencing, gates, process to work. Wooden doors that would not be hard to
R18 locks, safety/warning signs, inadequate break. One guard is on site during the night and is 56
security). responsible for both the WWTP and the WRP, but limited
possibilities to take action since he is alone and a rather
large area to cover. No locks on the chlorine station before
the sedimentation basin.
No final maintenance plan at the moment, still under
development. Supplier company maintains membranes. If
Failure due to inappropriate maintenance sedimentation basin is not maintained it will lead to more
R19 56
scheme (Not considering membranes). frequent backwashing. If no maintenance scheme you don't
know in which condition your installations are in, and
therefore more breakdowns can be expected.
Almost everything has some kind of backup system except
RO. The supplier maintains UF and RO once a month first
Failure of maintenance (e.g. No availability of
year. Shutdown of process in case of maintenance failure is
R20 spare parts, no knowledge of how to perform 56
noticed. Very limited availability of spare parts in Beaufort
maintenance) leading to failure.
West why long downtimes can be expected if spare parts
become necessary. |
Chalmers University of Technology | Risk
Hazard Hazard description Comments priority
number
If pre chlorination fails, less oxidation in sedimentation basin
Dosing malfunction due to equipment failure
leading to oxidization on membranes leading to more
R21 or power failure or improper dosing. Possible 37
backwashing. Too much will lead to shutdown due to
interruption of chlorination.
triggered by alarm before RO.
If the PLC cannot communicate with the on-line monitoring
the plant automatically shuts down resulting in process
R22 Failure of on-line monitoring system. 37
down time. This risk can therefore not be connected to
quality related risks.
The iron will lead to fouling of the membrane, needs one
R23 UF: Oxidation of iron on membranes. 37
day of backwashing.
Water temp affects RO process, algae growth increases
Problems related to low/high water or air during summer while the sedimentation basin needs more
R24 37
temperature. frequent cleaning, water temp affect the sludge age.
Increased risk of cyanobacteria.
If someone drops something in the cleanwater sump (only
wooden cover, not sealed) it may give quality effects since
there is no additional monitoring of this water before
distributed. RO and UF feed tanks are neither sealed, this
On-site reservoirs/ponds/watersumps can be
R4 may have quantity effects due to triggered alarms. It the 28
compromised/contaminated .
contamination is detected the plant needs to be shut down,
resulting in quantity problems. According to the contractor
the sumps will be sealed in the future to avoid accidental
contamination.
Trance membrane pressure alarm should pick this up, also
possible to noticed during maintenance. Possibility to repair
R6 RO: Membrane break trough; fiber breakage. smaller breakage on their own at site. Since this is final 28
membrane, quality may be affected if the alarm does not
work.
No final maintenance plan at the moment, still under
development. Membranes are maintained by contracted
Contamination or wear due to the use of company. If sedimentation basin is not maintained it will
R27 28
unsuitable materials. lead to more frequent backwashing. If no maintenance
scheme you don't know in which condition your installations
are in, and therefore more breakdowns can be expected. |
Chalmers University of Technology | Risk
Hazard Hazard description Comments priority
number
Don't keep stock of chemicals but chemicals are provided by
certificated trustable company (i.e. no chemical quality
R28 Chemical supply runs out. 28
problem). No water will leave the plant if there is a lack of
any chemical, therefore no quality effect, but quantity.
TMP-alarm should pick this up, otherwise it will be noticed
R29 UF: Membrane break trough; fiber breakage. 28
during maintenance.
UF: Increased membrane fouling, corrected Increased backwashing is leading to more downtime, but no
R25 24
with maintenance wash. other problems. Life length should not be affected.
Quantity consequences are referring to more backwashing
Inadequate monitoring resulting in inadequate
R1 as a result of inadequate monitoring of ingoing water quality 18
water quality or quantity risks.
(which may leading to un-optimized processes).
Non optimized process will lead to more backwashes, CIP
etc. Mainly a problem for the WWTP. This is where
Non optimized treatment processes can result optimization will have biggest impact. RO is not likely to fail.
R30 18
in poor process performance. Can treat almost any water to good quality, still there is a
risk if the membranes are not optimized. If not optimized
the UV is a final barrier for disinfection.
A lot of birds nearby the maturation pond where visible, and
a lot of bird dropping where confirmed on the water surface
Droppings of animals/birds (may e.g. introduce inside the pond. According to the operator the dropping will
R26 16
harmful micro-organisms into the water body) be removed in the sand filtration and the pathogens
removed in other barriers in the reclamation system. Due to
the origin of the water, this hazard should not be big risk. |
Chalmers University of Technology | ABSTRACT
DSM Food Specialties Seclin, unit of the International Group DSM, is leader in Enzyme Production. Because of
the sensitivity of these products and its batch production, a high process control is needed. In that sense the site
has implemented a Tool called OPN-AU Tool to monitor important parameters during critical operations and
analyze productivity, costs and bottlenecks of the plant production.
However, the configuration of the Tool needed many improvements and a project was on runaway to include a
new feature to computerize the Shift Reports actually performed by Operators on paper sheets. Thus, this
Master‟s Thesis aimed to increase the efficiency of the Tool and participate in the management of the new project.
Main problems notified on the existing configuration were the absence of monitoring for new enzymes and
maladjusted settings which resulted in aberrant alarms on the plant. Indeed, when a drift of parameters is pointed
out, Operators are informed by red or orange alarms but their no relevant activation implies a decrease of interest
regarding the Tool. Production Staff did not trust it anymore and did not take care anymore about alarms.
Initially, the study was focused on learning accurately the process performed in Seclin which includes many
different operations (fermentation, extraction, filtration, concentration, formulation, drying, packaging…). Then
many meetings with Operators, Experts and Engineers have been done to understand current problems on the
Tool and its configuration. Finally, modifications have been implemented and checked directly on the plant in
order to fix issues and to improve what was needed.
This Master‟s Thesis was successful by having reached almost all its objectives. Shift Report Tool is now ordered
after having negotiated its price and its Functional Design Specification. It will be implemented and tested on the
plant in June. Even if some improvements are still needed, the OPN-AU Tool‟s efficiency has increased
considerably: In four months the number of aberrant alarms decreased from 40% to less than 20% and three new
products have been entirely configured on whole the production line. It was a success as well on a human aspect
since many training sessions and meetings with Production Staff have been performed. It assured a rapid detection
of problems, a rekindled enthusiasm regarding the Tool and so the return of an active participation from all in its
improvement.
Key words: Enzyme production – Process Control – Configuration – Optimization – Standards – Alarms.
Configuration and Optimization of a Process Monitoring Tool 1 |
Chalmers University of Technology | ACKNOWLEDGE
First, I would like to express my gratitude to Mr. Per-Henrik Larsen, Chief Executive Officer, and Mr. Hervé
Denoncin, General Director, for welcoming me in DSM Food Specialties Seclin.
I wish to thank Mr. Fabian Flocard, Drying and Packaging Technical Expert, especially for the confidence he has
expressed to me, his daily coaching and the opportunity he gave me to discover a company which has an
international dimension. It was a real pleasure to work with him and I really appreciated the freedom of action and
decision he let me. I guess his plant will remember my capacity of taking initiatives.
I also thank Ms. Magali Ricarde, Teacher Researcher at ENSGTI in Pau (France), and Mr. Anders Rasmuson,
Teacher Researcher at Chalmers University of Technology in Gothenburg (Sweden), as tutors of this Master‟s
Thesis, for their availability and their advices despite distance which did not help.
I would like to thank as well Mr. Nicolas Legrand, Recovery Expert, Expert Leader, Golf Champion and so
serious opponent of Tiger Woods, for the share of his knowledge especially about DSP sector and fuchsia fashion.
It was a really pleasure to share the office with him and Mr. Fabian Flocard even if it has to be said that some
improvements are needed in the ball launching.
All my gratitude goes to Mr. Francis Maryniak, Fermentation Expert, Mr. Jean-Claude Cardon, Chromatography
Expert, and Mr. Paul Manducher, Production Manager, for their availability, their advices, their support and their
sympathy which let me to work in a good atmosphere.
Finally, I wish to thank all Staff of DSM Food Specialties Seclin, in particular all Operators for the time they
spent answering my questions, their hospitality and their kindness. It let me to evolve in a pleasant way in this
male environment.
Configuration and Optimization of a Process Monitoring Tool 2 |
Chalmers University of Technology | INTRODUCTION
Important actors in Food Industry, Enzyme producers have a booming market for few decades. The Dutch
International Group DSM has a unit leader in this field: DSM Food Specialties.
This Master‟s Thesis was conducted during a period of six months in France in DSM Seclin, a production plant of
this unit. Each year, the site produces around 8.000 tons of enzymes from various micro-organisms. Developed in
Fermentation, they undergo several steps of extraction and concentration before being packed.
DSM understands that the key to success lies in an optimization of operations to get a maximum performance
from his equipment and assure a quality of produced enzymes unmatched. In this context, my internship was
devoted to configuration and optimization of a Tool, OPN-AU Tool, whose goals are the establishment of process
monitoring and the standardization of manufacturing processes. It included as well the management of a new
project, the creation of Shift Report Tool which should allow computerization of teams‟ reports.
The task was therefore to improve the process control, its analysis and to perform project management. It fits
perfectly with the educational goals of a Master‟s Thesis. Indeed, these tasks put the student in a situation of
professional responsibility since they imply learning of scientific, technical and human resources in the
engineering profession and thus prepare his employability.
This report is divided into three main parts. First, focused on generalities, it presents the company and the
description of manufacturing processes of enzymes. Then it details means to control the process on the site,
including the configuration and the use of OPN-AU Tool. Finally, it describes the main concrete actions carried
on during this mission through various examples and their impacts on production.
Configuration and Optimization of a Process Monitoring Tool 5 |
Chalmers University of Technology | GENERALITIES
A. General presentation of the company
1. DSM Group
Dutch States Mines (DSM) is an international Dutch group which creates
innovative products and services in Life Sciences and Materials
Sciences, contributing to the quality of life. DSM‟s products are used
globally in a wide range of markets and applications, supporting a
healthier, more sustainable and enjoyable way of life. This company,
located on five continents, is present in more than 49 different countries
as shown on Figure 1. Indeed, it is worldly known as supplier of the
chemical industry. It employs more than 24,000 people worldwide and
with a turnover of almost EUR 8.8 billion it is even listed on Euronext
Amsterdam. [1]
Figure 1: Localisation of DSM sites [1].
Initially, DSM was focused on the coal market but nowadays, as shown on Figure 2, it is more diversified and
includes human and animal nutrition and health, personal care, pharmaceuticals, automotive, coatings and paint,
electrics and electronics, life protection and housing.
Figure 2: Market of DSM [1].
2. DFS Group
DSM Food Specialties (DFS) is part of nutrition cluster of DSM. It is a global supplier of advanced ingredients
for food and beverage industries, primarily manufactures with the aid of fermentation and enzyme technology. Its
products contribute in a major way to the success of the world favourite dairy, processed food, soft drink, fruit
juice, alcoholic beverage and functional food brands.
Its headquarters is located in Delft in Netherlands. DFS
represents more than 30% of the DSM turnover and has
the higher number of employees. This department
employs more than 1,600 people spread in 20 business
agencies, 3 Research and Development sites and 10
production factories. [1]
More than 50 different products are sold around the
world over more than 1,000 formulations. Indeed, as
shown on Figure 3, DFS is composed of different
business groups.
Figure 3: Structure of DSM Food Specialties [1].
Configuration and Optimization of a Process Monitoring Tool 6 |
Chalmers University of Technology | These various specialties let DFS to have a large market position in several areas:
Dairy
Savoury
Baking
Beverages
Functional Food
Enzymes
With its two manufacturing plants (France-Seclin and Italy-Lavis), DFS Enzymes performed to have a presence
all around the world (Figure 4).
Figure 4: DSM Enzymes, a worldwide presence [1].
3. DFS Seclin
a. History of DFS Seclin
DFS Seclin has been a pioneer of industrial biotechnology. It began in 1906 when Auguste Boidin discovered an
amylase produced by bacteria. Because of its capacity to accelerate the liquefaction of starch, it had been called
Rapidase. Thus, in 1922, August Boidin and Jean Effront built a first industrial production of enzymes in Seclin:
Company Rapidase. Then, as described on Figure 5, the company took different names through last decades. It is
in 1998 that it became DSM Food Specialties Seclin after a merger with the second chemist Dutch DSM. [2]
Figure 5: History of DSM Food Specialties Seclin [2].
b. DFS Seclin Nowadays
Activity
DFS Seclin is specialized in manufacturing and marketing of functional enzymes for Food industry. Thus, its
enzymes are present in many food ingredients and contribute to every day food improvement.
For instance, they can improve products quality (volume, texture,
appearance…) and process quality (machine ability, dough
tolerance), increase the yield or reduce the production cost as well.
That is why they are used in various sectors such as wine industries,
food, baking and brewing processes such as the Karmeliet
production (Figure 6). [2]
Figure 6: Sectors improved with enzymes application [2].
Configuration and Optimization of a Process Monitoring Tool 7 |
Chalmers University of Technology | Structure
This production plant is located in Seclin (France-59) on a site of 2.3 ha for more than 180 employees (Annexe I).
It was certified ISO 9002 by AFAQ in 1993, 1996 and 1999 and certified by the new ISO 9001 in 2003 and 2006
[3]. Moreover, it is currently on a process of AIB Certification.
The Site Director is assisted by the Management Committee (CODIR) which each member is responsible for one
or more services comprising managers, supervisors, technicians and/or operators according to its structure
(Annexe II).
The organization is based on the cooperation of several departments such as production, the engineering
department, quality service, the production support (including Maintenance Services, Environment Unit…), the
business unit and the communication department.
The production department is mainly composed of four areas directed by an Expert, specialist of it:
Fermentation
Down Stream Processing (DSP)
Chromatography
Finishing (Drying and Packaging)
The plant operates as a batch production 7 days 7 and 24 hours a day in “3/8” (i.e.: Organization of work where 3
shifts work 8 hours each to cover 24 hours a day). Thanks to this no-stop production, DSM Seclin manages to
produce more than 7,600 tons a year including 6,200 tons in liquid form and 1,400 tons in solid form [2]. It has
been decided to do a batch production because of the diversity of the products. Around 25 different enzymes are
packaged in 20 types of packaging, marketed under nearly 200 different formulations which represent more than
400 references in the final active packages. It is around 80% of its production which guarantees its presence all
around the world. [4]
B. Process Description
Enzyme is an active protein which makes it possible to transform a complex natural substance into simpler and
more easily absorbable substance [3]. The industrial ones produced in Seclin are made by fermentation of micro-
organisms. It can be done with bacteria, yeast or mold.
Figure 7: Fermented media of Yeast (A), Mold (B) and Bacteria (C) [3].
The process consists in the multiplication of these micro-organisms inside the media which synthesize enzymes in
big quantity. It is necessary then to extract only the enzymes and perform packaging (Figure 8). That is why the
production process is divided into four main areas:
Fermentation
Extraction and filtration
DSP Phase
Formulation
Finishing (Drying and Packaging)
Configuration and Optimization of a Process Monitoring Tool 8 |
Chalmers University of Technology | 1. Fermentation
Fermentation is the growth of micro-organisms in vessels of various sizes, from a laboratory scale to an industrial
one. It is divided in six different units as shown on Figure 9:
Figure 9: Steps of the fermentation phase [2].
All along these steps, the capacity of equipments is increased. The purpose of this scaling up is to decrease the
duration of the latent phase in the micro-organisms growth (Figure 10) so as to have a minimum occupation of
facilities for a maximal production of enzymes. Indeed, during this phase, a micro-organism does not multiply and
does not produce anything, it just adapts to the environment. [5]
However, even if the fermentation phase is divided in six steps because of the needed scaling-up, they represent
actually two different stages:
The biomass growth, from the production strain step to the inoculum step.
The main fermentation for the enzyme production.
a. Biomass Growth
This first stage is performed in the microbiologic laboratory. The strains selected and
prepared by the headquarters at Delft, are sent to Seclin as spangle where they are first
revived in order to place them under optimal conditions for their development. In that
purpose, strains are mixed with a culture medium where mixing, temperature and air are
controlled. As shown on Figure 9, biomasses grow first in a tube, and then it is transferred to
a flask in order to be finally put in a Punt-Bus.
Figure 11: A Punt Bus.
Once strains are mature and so completely developed inside the Punt-Bus (i.e.: it has a
sufficient biomass concentration), they are seeding to an inoculum which has been previously
sterilized and filled of a culture medium. Thus, micro-organisms use organic and inorganic
materials available in it in order to multiply and synthesize proteins, including the wanted
enzymes.
Figure 12: An inoculum.
b. Enzymes Production
It is essentially during the main fermentation stage that enzymes are synthesized. As in the
case of an inoculum, a fermentor is first sterilized and filled by a culture medium before
being inoculated (i.e.: filled with the inoculum content).
Both inoculum and fermentor are working as batch and feed-batch processes [3]:
Figure 13: A fermentor.
Batch: Before the seeding and the inoculation, the tanks‟ bottom is filled with a medium which contains
all nutrients (glucose…) needed for the development of strains and enzyme production.
Feed-Batch: After a while, nutrients initially present in the medium are no longer sufficient to ensure
adequate growth of strains. Therefore, it is necessary to feed the fermentor with other nutrients from a
smaller tank called Satellite. For each fermentor, there are three satellites respectively filled by food
(carbon source: sugar, nitrogen source: salt or flour, minerals, vitamins…), acid or base and anti-foam.
Configuration and Optimization of a Process Monitoring Tool 9 |
Chalmers University of Technology | Because of this addition of products, regularly it is necessary to do some withdrawals sent to other vessels in
respecting sterility of equipments where they will be treated (extraction step). Sterility and purity are really crucial
in this kind of enzymes production for two main reasons:
It involves the development of micro-organisms. A presence of contaminant could imply a phenomenon
of competition and therefore a loss of yield.
Enzymes produced in DSM Seclin are used in Food industry, so the final product has to not contain any
harmful contaminant.
In that purpose and to insure an optimum growth of micro-organisms, enzymes production requires control of the
conditions in the tanks like air, pH, temperature and purity. That is why several parameters are monitored [3]:
Aeration: It supplies the amount of substrate oxygen needed for the micro-organism‟s respiration. The
rate of oxygen consumption is reflected in the OUR (Oxygen Uptake Rate). This measure represents the
amount of oxygen consumed per kilogram of mash per hour. It is expressed in mM / kg / h.
Counter-Pressure (CP): The oxygen supplied by air is not very soluble in water. It is therefore not
available for micro-organisms. To increase the dissolution, CP must be increased (i.e.: higher increase in
oxygen saturation).
Mixing: It also increases the dissolved oxygen by shearing of air bubbles which increases the exchange
surface.
DOC: It represents the amount of oxygen dissolved in water. In this process, when the DOC (Chemical
Demand of Oxygen) is less than 30% of its initial value, it is assume that oxygen becomes limiting.
Temperature: It is maintained at a set point through cooling water (fermentation is exothermic).
pH: It is maintained at a set point that varies depending on the strain. The regulation is done by ammonia,
soda, sulphuric acid or phosphoric acid. Its measurement is made from sterilized and pressurized glass
sensors.
Once fermentation is over, the content is added to vessels where it will be treated and prepared for the filtration
phase.
2. Extraction and Filtration
The extraction phase is a filtration to extract enzymes from fermentation broth and to eliminate the biomass and
the residual part of raw materials. This takes place in different stages as described on Figure 14:
Figure 14: Steps of the extraction phase [2].
a. Preparation Phase
Before performing any kind of filtration the fermented medium from fermentors has to be treated. Indeed,
generally the suspension is broken down into two fractions:
A soluble one containing soluble enzymes, minerals, and organic matter.
An insoluble one containing residual cells and waste materials.
Configuration and Optimization of a Process Monitoring Tool 10 |
Chalmers University of Technology | It is the case for most of enzymes produced on the plant: They are extra-cellular and so are dissolved in the
culture medium. However, some of them are intra-cellular and so require an autolysis of micro-organisms in order
to be released (Figure 15). It is the killing phase which is usually made with the addition of chemicals (e.g.
octanol), a high temperature and a low/high pH. [3]
Figure 15: Medium with extra-cellular enzymes (A), with intra-cellular
enzymes (B), with intra-cellular enzymes after an autolysis (C) [3].
Moreover, whole of cultivated strains have to be destroyed
before discharging to environment in order to respect law. This
treatment is always accompanied by addition of chemicals and a
temperature kept equal to a certain value during a given
duration. [6]
After this optional killing phase, the wort is prepared by adding
an adjuvant, improving the filterability and, in some cases, CaCl
2
in order to reduce the problems linked to the presence of organic
acids (Figure 16). It is the harvesting phase.
Figure 16: Flocculation principle in a fermented medium [3].
Once these operations are done, the filtration and extraction phases can begin.
b. Clarification Phase
First, the product is clarified by using membranes filters press (FAM) in order to
eliminate the insoluble matters and especially the biomass. That lets to have enzymes
in a clear liquid (Figure 18).
Figure 17: A membranes Filter press.
This filtration takes place in several cycles themselves broken down into several steps [7]:
Filling: Application of a thin layer of additive on filters in order to improve the filterability.
Filtration: Retention of insoluble materials by feeding at a high pressure.
Pushing: Cleaning of pipes.
Compaction: Compression of formed cakes to extract the maximum of enzymes and facilitate the
removing of cakes.
Washing: Water flow in counter-current to recover the maximum amount of solute.
Removing: Cakes are removed.
Figure 18: Clarification phase [3].
Configuration and Optimization of a Process Monitoring Tool 11 |
Chalmers University of Technology | c. Plate Filtration Phase
After the clarification phase, the product comes from the membranes filters press to storage vessels where are
done some pH corrections and preparations for the subsequent step of the DSP phase, the plate filtration phase.
The purpose of these preparations is to avoid clogging plates during filtration. This step is
used to remove small particles and possible contaminants. It is a sterilized operation where
once again, it is the filtrate which is recovered. Thus, the filter (FAP1) consists of two parts
with different porosity of plates; a pre-filter with a higher porosity and another filter with a
smaller porosity to minimize contamination. Three phenomena have to be considered at this
level [7] [8] [9] [10]: Figure 19: Plate filter.
Surface filter that traps particles on the surface of the plate.
Depth filtration which retains more fine particles in the pores of the plates.
Adsorption filtration which removes negatively charged particles.
Figure 20: Phenomena of filtration on a plate filter [8].
After that, another phase of storage and preparation is performed in order to prepare the next step of the DSP
phase.
d. Concentration Phase
Once the preparation done, the purpose is to concentrate the enzyme solution. It can be done in two different ways
[9] [10]:
Ultrafiltration where water and mineral salts are
eliminated in permeate by passing it through the
membrane pores.
Figure 21: Ultrafiltration principle [3].
Chromatography used for certain products (mainly for the dairy industry) where purity has to be ensured.
It operates on the ion exchange principle and hydrophobic interactions. It is coupled to a final
concentration by ultrafiltration as well.
Figure 22: Chromatography. Figure 23: Ultrafiltration.
At the end of these extraction phases, the enzymatic concentrates are stabilized (by adding salt, glycerol) to
reduce the risk of contamination and loss of enzyme activity during storage. They were finally stored at low
temperature and controlled pH before being sent to the formulation phase [9].
Configuration and Optimization of a Process Monitoring Tool 12 |
Chalmers University of Technology | 3. Formulation and Packaging Phase
This last phase of the production line is composed of different steps as shown on Figure 24.
Figure 24: Formulation and packaging phases [2].
a. Formulation Phase
This step is used to make products fitting with the specifications needed by customers. Thus,
it may induce the addition of more concentrated enzyme, salt, water or a pH correction. The
product passes then by a second plate filter (FAP2) in order to secure high quality of the final
product. It includes the elimination of eventual precipitates or possible contaminants so as to
correspond to the specifications of Food Industries. Finally, the finished product is stored
awaiting the packaging step.
Figure 25: Formulation
tanks.
b. Packaging Phase
DSM Seclin proposes to its customers enzymes under liquid and solid phases. Thus, the packaging phase, as
described on Figure 24 is divided in two different areas: the liquid packaging and the solid one.
Liquid packaging
Products which have to stay under a liquid phase are packaged just after the
formulation phase. It can be done in containers or drums.
Figure 26: Liquid packaging line.
Solid packaging
Product intented to be marketed in solid form have to be transformed into granules in a dryer.
That is performed with a sequence of different processes [7][8][9][10]:
Drying: The liquid is spread in fine droplets through high pressure nozzles under hot air
conditions (convection drying).
Fluidized bed: To finish the drying and sorting particles.
Cyclone: Smallest particles are recovered in a cyclone. Once they are agglomerated they
are sent back to the dryer.
Figure 27: Solid
Finally, the resulting product is mixed, standardized and then packaged in BigBags, pots or boxes. Packaging.
4. Quality Aspects
All by-products and waste generated during the process are treated or reused. Solids are incinerated or directly
spread on fields and liquids are treated in sewage treatment plant. DSM Seclin has its own water treatment plant.
This facility allows it to maintain compliance with legal requirements on the disposal of sewage. It was built at
1.5 km from the production plant so it is underground pipes which send effluents to the station.This sewage
treatment plant is expected to treat 4,000 m3 load a day. [4]
In addition to some checks of purity, enzymatic activities are systematically controlled at different steps of
production. The main purpose is to ensure that the final product perfectly fit with the clients' expectations but it
also lets to quickly detect stages which result in a low yield.
Configuration and Optimization of a Process Monitoring Tool 13 |
Chalmers University of Technology | In the same way, majority of the production line is done over steril and so clean conditions. Because it is a batch
process, it implies a cleaning of each equipment after each use. That is performed by two different ways both fully
automated [4]:
“Cleaning In Place” Station (CIP): There are three CIP stations, each dedicated to different facilities. The
cleaning solutions are made up of acids, bases and disinfectants. They are successively used and projected
by spray balls.
Included Cleaning: Membrane filters press, dryer, mixer, chromatography and ultrafiltration units have
their own cleaning stations which are actually an integral part of these facilities.
To facilitate the process‟control and so for quality and bussiness aspects, all the process line is monitored
informatically. Indeed, a monitoring and supervising system manages all vessels and certain facilities in order to
control and regulate many parameters, make the transfers, etc. It is performed by several automatons and the main
one is ABB.
Configuration and Optimization of a Process Monitoring Tool 14 |
Chalmers University of Technology | PROCESS MONITORING
A. Generalities
The overall objective for DSM is to become a reliable supplier of high quality products at competitive prices in all
sectors that it covers. Thus, its strategy revolves around three key drivers:
Growth: Driven by markets and innovation.
An increased presence: Especially in emerging economies.
Operational excellence: Cost optimization and profitability.
In this sense, few years ago, DSM Group introduced on its production plants a standard called MANUFEX
(MANUfacturing EXcellence).
1. MANUFEX
As part of the operational excellence point, MANUFEX aims to optimize performance, improve reliability of the
site and so optimize all steps of a process (production, logistic, business, maintenance...). Indeed, it should assure
a continuous improvement of whole of operations but as well a good opportunity to staff for increasing their
autonomy and their ability to take decisions by themselves. As described on Figure 28, MANUFEX is composed
of several Work Processes relating to the process, maintenance, research and so on. [4]
Figure 28: MANUFEX composition [4].
Three main Work Processes are relevant for the production area:
Manufacturing Execution System (MES)
Operate Plant Normal (OPN)
OPN-AU Tool
Asset Utilization (AU)
2. MES Interface
As soon as the customer service of DSM receives a new order of a product made in DSM Seclin, it sends it to the
sales of DSM in Delft, headquarters of DSF. In consequence, the sell order is placed in the list of productions for
Seclin where it will be then automatically processed in a planning order. The inventory is checked and if the
product is no longer present in the stock, a demand for fermentation production is generated and so the planning
order is automatically transformed into a Process Order (PO). [11]
The product is then integrated into the detailed planning. Since it is a batch production, Seclin planners check and
adjust if necessary planning or tanks used in the light of information received by the production area (vessels not
available, equipment failures ...).
Configuration and Optimization of a Process Monitoring Tool 15 |
Chalmers University of Technology | Few days before starting fermentation, production administration initiates the preparation of raw materials needed
for manufacturing and releases the order of preparation which will be sent to MES. Thus the inspection plan for
this product is created: the control laboratory may enter data on current production as soon as the tests are
available. [11]
In consequence, the process order is visible to Operators of each workshop of the factory. Whenever it is
stipulated in the “Code 4” (i.e. procedures that operators have to followed during the production), Operators have
to activate on MES a Process Order (PO) on the equipment they used. It can be comparable to having folders for
each device and writing actions that Operators made. (Figure 29)
Once a PO is open, MES knows it has to register all data relating to the product and
equipment. For instance, additions of raw materials from a network (glucose, water...) are
automatically recorded by MES by means of various measuring equipments such as flow
meters, counters or weigh tanks. These data come from automatons used on the plant such as
ABB. A Tag is assigned for each parameter followed on ABB; it is used as a sequence which
lets to identify sensors‟ values. Thus, these Tags are used in every informatics database to
express specific parameters (pH, Temperature…); ABB assigns to a Tag a value which is
transferred to MES and updated on line. Figure 29: PO Opening.
However, manual additions made by Operators are not automatically saved. In consequence, Operators have to be
actively involved and fill them by themselves in MES. Moreover, the occupation time of a resource is recorded.
Once raw materials are added, MES registers the various consumptions which automatically updates the
inventory. As a result, stocks are always updated.
In order to follow production in a right way, POs are really specific for
each equipment and even each process which takes place on it. So on
MES are configured each equipment and each tank of the plant. Then
Operators have to specify on which facility they are working on (Figure
30).
In that sense, there are lots of different kinds of POs and each PO is
composed of different phases. These ones can be selected automatically
by the system when Operators do specific things such as opening a valve
or it can be done manually by Operators. [11] Figure 30: MES Interface- Choice of Equipments.
For instance, during fermentation, a PO is open on the fermentor used: PO Fermentation.
Representing the process, this PO is actually divided into three different steps [12]:
Preparation: When the fermentor is sterilized and filled by the culture medium.
Fermentation: When fermentation takes place.
CIP: When fermentation is over and the fermentor is cleaned with a CIP station.
On MES, transfers from one PO/phase to another one are automatically done when a specific condition is filled as
for PO Fermentation described on Figure 31:
Figure 31: PO fermentation description [12].
This aspect of MES is really important for this study because the configuration of OPN-AU Tool relies on it.
3. OPN-AU Tool
MES allows a new and performance-oriented process management. Thus optimal operation regarding capacity,
yield, quality, time… is achievable with product-specific parameters like KOP`s (Key Operating Parameter),
KPP's (Key Process Parameter) and KPI's (Key Performance Indicator). These parameters can be monitored
thanks to OPN-AU Tool.
Configuration and Optimization of a Process Monitoring Tool 16 |
Chalmers University of Technology | For the moment, it is composed of two tools:
OPN (Operate Plant Normal): This part of MANUFEX program aims to formalize the
establishment of a process systematic and standardized monitoring and so an optimization of
P production steps. This check is done through parameters (Temperature, pH…) followed in real time
O N
which lets to evaluate the production performance. So it lets to identify gaps and excesses in
production operations, to diagnose the cause of the latter and to determine appropriate corrective
actions.
AU (Asset Utilization): Its prior goal is the online comparison of the present performance of the
production with the possibilities of the line and its resources. Hence it is possible to clearly
determine the potential of optimization regarding capacity, plant utilisation and yield. AU Tool is
ready to identify cost drivers and performance killers via the definition of performance targets.
A project is underway in order to add another tool to OPN-AU Tool. It will be called Shift Report Tool and will
be used by operators to do their Shift reports on computers instead of doing it manually on a paper sheet. Several
fields will be directly filled from MES and OPN database which should let to save time, to decrease the number of
human mistakes and so to optimize the productivity of the plant. In consequence, in few months OPN-AU Tool
will become OPN-AU-Shift Tool.
These Tools will be more detailed in the following parts but one can already said that the strength of OPN-AU
Tool is its adaptation to the complexity of the plant including a large flexibility, a batch processing and a very
wide range of products. That goes in the sense of the MANUFEX program‟s purpose: Becoming leaders in
Enzyme production. [4]
B. Process Control
1. OPN Tool
a. Generalities
OPN Tool is used in order to monitor accurately batches. In concrete terms, for each step of each production Key
Operating Parameters (KOP) are defined. These parameters are the ones considered as the most important, most
representative and influential on the performance of a specific operation and quality of final products. Selected in
small numbers, they can be more easily accessible and able to bring greater attention to follow-up. [13]
For instance, at DSM Seclin, pH and temperature are systematically defined as KOPs for all stages of production
because of the enzymes sensibility. However, some others are added depending of the performed operation. The
KOPs listing is decided for all facilities and all different products by Experts and Members of the Scientific
Production Staff (SPS).
Then, as described on Figure 32, each KOP is associated with Operating Windows (OW) which let to compare the
actual value of a KOP to:
An optimum.
A green area where the operation is under control.
An orange area where the operation is experiencing a drift and requires some precautions.
A red area where the operation has a significant drift and requires immediate corrective actions.
The purpose of these OWs is to have a visual tool to quickly detect a problem and be reactive or proactive when a
KOP drift occurs. [13]
When a drift of production happens, some procedures are put in place on OPN Tool to guide Operators who have
to rectify it. These are called corrective action trees and can be considered as a capitalization and a standardization
of best practices to implement when some discrepancies occur on KOPs‟ values.
Configuration and Optimization of a Process Monitoring Tool 17 |
Chalmers University of Technology | These corrective action trees are classified into two categories [13]:
Preventive measures: Implemented when KOP goes into the orange area, they can overcome a
number of checkpoints that may be the source of the noticed gap.
Corrective actions: Implemented when KOP goes into the red area, they limit impacts of a serious
abuse of the manufacturing process.
Annexe III shows an example of a corrective action tree when temperature is higher than what it should be during
the fermentation Phase.
In consequence, as soon as a KOP drift is noticed, an action can be taken by Operators without wasting time by
directly calling Shift Leaders or Experts. It assures a high process control and a fast reaction capacity.
b. Advantages
OPN Tool improves the process control in different ways. Indeed, it lets to:
Insure a more efficient production.
Get a better products‟ quality.
Standardize procedures to control production‟s steps.
Thus, it generates benefits both on a human and a technical aspects.
Human aspect
OPN Tool is a good way to empower Operators. It limits the number of elements to monitor so Users can be more
reactive; they can anticipate problems and try to fix them by themselves thanks to the corrective action trees. Thus
they have a higher force decision making and a wider field of action. They have more control on their operations
and they can actively participate in problem solving and share their ideas to assure a continuous improvement.
[14]
In consequence, more than implying a better monitoring of the production line, OPN-AU Tool lets Production
Staff to take initiatives and so feel more concerned by the running of the production.
Technical aspect
An optimization of the production can be performed thanks to OPN Tool because it lets to do an overall
assessment of each operation and so to analyze and define steps which present problems or abnormal low results.
This feature is called: Scoring.
When Experts and SPS members defined with Administrator the KOPs listing for an Enzyme A on given
equipment, they confer a weight factor to each KOP proportional to its weight of its importance along the process.
Let‟s take the example of Enzyme A through the inoculum phase process where KOPs are pH, temperature and
feed rate. The product is more sensitive to temperature than to pH and feed rate. So the weight factor assigned to
temperature is higher than the ones set for other KOPs (Table 1).
Table 1: Weight factors of KOPs on a inoculum for Enzyme A.
KOP Weight factor
pH 10
Feed Rate 10
Temperature 20
Configuration and Optimization of a Process Monitoring Tool 18 |
Chalmers University of Technology | Once the operation is done, for each KOP, the percentage of numbers of read values which are included in the
green area, the orange one and the red one can be calculated. So a mark can be attributed to each KOP. It is equal
to the sum of percentages of green and orange areas time the weight factor of the KOP.
For example, in Table 2:
Mark = (green % + orange %) x KOP weight factor.
Temperature Mark: (5% + 20%) x 20 = 5/20.
Table 2: Marks of KOPs for Enzyme A on the inoculum phase.
KOP Weight factor Green (%) Orange (%) Red (%) Mark
pH 10 50 50 0 10/10
Feed Rate 10 0 50 50 5/10
Temperature 20 5 20 75 5/20
Then, the sum of all marks can lead to the Scoring of the operation.
By these calculations, KOPs which had some wrong impact on the process can already be pointed out.
In the example, the Scoring is equal to: (10+5+5) x 100 / (10+10+20) = 50/100.
Thus, it can be deduced that:
Temperature got quite bad values:
Operators should be careful to it for the next operation.
The Scoring is quite low:
If the operation had a high yield, the KOPs list or KOPs‟ weight factor may be improved. So
correlations are made by SPS members to check that KOPs and KOPs‟ profiles choices are
relevant.
To define the yield, two parameters are needed:
For each product is calculated a unit of capacity measurement on each fermentor. It is called Deca
Hectoliter Week (DHW) and is defined as below:
After each operation on the plant a test to define the enzymatic activity of the product is made and is
expressed in BU (Billion Units of Enzymes).
The yield, expressed in BU/DHW, is usually calculated as the ratio of the Enzyme activity over the DHW
number.
Then the correlation between the KOPs list and the obtained yield (BU/DHW) is checked by SPS members who
study curves representing yields obtained on same product‟s different batches versus Scoring.
For instance, Figure 33 represents fermentation yields obtained from Enzyme A Batches during a given period
versus the resulting fermentation Scoring. With a correlation factor equal to 0.75, it seems that the choice of the
KOPs is relevant. [15]
The same thing may be performed to study the impact of operations on a sector which let to define steps causing
main problems. It determines on which operations people should focus on to improve the production. Each
operation has a weight factor, exactly as KOPs, and from each Scoring can be deduced Global Scoring of an area
of the production or even Global Scoring of the plant. For instance, Figure 34 represents the calculation of the
Scoring got for a specific batch on the DSP phase. From the plot on the right can be deduced that E4 is the main
equipment where abnormal results can be noticed.
Configuration and Optimization of a Process Monitoring Tool 19 |
Chalmers University of Technology | In consequence, more than a monitoring Tool, OPN lets Operators to be more independent and autonomous and
can result in a real improvement of the production thanks to the Scoring studies. OPN Tool allows Managers and
Operational Experts to have an analysis tool more efficient to identify weaknesses and failing facilities that imply
abnormal low results. Thus, they can more easily point out elements which optimize the production control and so
determine rapidly the relevant research areas. In one word, this Tool makes possible to quickly define the limiting
steps of production which can be improved.
2. AU Tool
a. Generalities
AU Tool aims to determine the maximum production capacity that the plant can obtain and compare it to each
specific performance actually achieved.
Three different studies can be performed on AU Tool to calculate Opportunity Gaps (OGAP) between real
measurements obtained on the plant and standards defined by Business Units or Experts[13]:
AU Time
AU Result
AU Occupation
AU Time
Such as parameters defined as KOPs on OPN Tool, are defined standards which can be considered as Key
Performance Indicators (KPI):
Delay KPI: It determines the delay between the actual start of an operation and the moment when it should
have started according to the planning.
Duration KPI: It compares standard duration of the operation with the real time that has been needed.
They are configured exactly in the same way than KOPs on OPN Tool (per product and per equipment). As
shown in Table 3, these two KPIs (expressed in minutes), are then multiplied by the cost of one minute of the
equipment‟s occupation in order to define OGAPs. Thus, AU Time is a good way to see the cost of possible
delays on different operations and so the potential bottlenecks of the production line. [13]
Table 3: AU Time example.
Equipment-Product Scheduled Start Real Start 1 min Cost OGAP Delay
Inoculum-Enzyme A 8:00 am 9:00 am 50 €/min 300 €
Equipment-Product Standard Duration Real Duration 1 min Cost OGAP Duration
Inoculum-Enzyme A 100 h 150 h 50 €/min 150 k€
AU Result
This function of AU Tool aims to study possible OGAPs regarding yields of each operation on the plant. In
concrete terms, the actual yield is compared with the Maximum Proven Capability (MPC). This parameter is a
standard defined as the third best yield already obtained during the operation. So its value is really variable. [13]
Table 4: Study of a yield.
Equipment-Product Yield MPC Gap
Inoculum-Enzyme A 75 BU/DHW 80 BU/DHW (80-75)x80/100 = 4 %
The Gap is then transformed in a DHW Gab by multiplying it with the DHW number related to the enzyme and
equipment used. Finally, as done in the AU Time, it is multiplied with the cost of 1 DHW and it gets the OGAP.
Configuration and Optimization of a Process Monitoring Tool 20 |
Chalmers University of Technology | Table 5: AU Result Example.
Equipment-Product Gap DHW DHW Gap 1 DHW Cost OGAP
Inoculum-Enzyme A 4 % 2 4x2 = 8 DHW 100 k€/DHW 800 k€
Thus, AU Result is a good way to define the cost generated by obtaining low yields.
AU Occupation
Based on POs, AU Occupation is used to determine the cost of inoccupation of equipment. In that sense, it
defined the duration that equipment is not used during a given period. This number multiplied the benefit of one
hour of production gives the Occupation OGAP. [13]
Table 6: Study of a yield.
Equipment End of PO 1 Start of PO 2 Inoccupation Benefit/min OGAP
Inoculum 8:00 am 9:00 am 1 h 100 €/min 6 k€
Thus, AU Occupation is a good way to define the cost of unuse of equipments.
b. Advantages
Thanks to AU Tool, the evaluation elements present on OPN can be linked to the budgetary aspect of the
production to determine the limiting and not optimal steps. In consequence, this study helps to identify
opportunities for improvement associated with specific functions. The recurrence of these missed opportunities
allows a ranking of "Performance Killers" or issue limited according to their importance. It is then easier to focus
on the key and work actively to fix problems causing losses of time and productivity. It is really successful in a
batch-oriented environment, where necessarily a high resolution of production steps (phases, runtimes etc.) is
demanded, to detect any process drift as soon as possible. [13]
OGAPs‟ calculations can be then analyzed through three different graphs:
Gap per Product
Gap per Batch
Top 10 Performance Killers.
Gap per Product
This graph represents OGAPs versus Products. OGAP for one product can be the
sum or the average of OGAPs obtained on different batches. Performed by
Experts, this analysis aims to define which product implies the highest OGAP
and so the most important problems on the production line. For instance, on
Figure 35, Product D seems the one which implies the highest OGAP. Thus,
Staff should first focus its study on it to improve as much as possible the
productivity.
Figure 35: Gap per Product plot.
Gap per Batch
This graph represents OGAPs versus Batches of a product A performed during
a given period. Performed by Experts as well, it aims to define which batch
implies the highest OGAP and so the most important problems on the
production line. That is really useful for example when SPS members try some
changes on the production procedure. For instance, on Figure 36, it seems that
Batch 5 was the less efficient.
Figure 36: Gap per Batch plot of a given
Enzyme A production.
Configuration and Optimization of a Process Monitoring Tool 21 |
Chalmers University of Technology | TOP 10 Performance Killers
Some standards are defined according to OGAPs‟ values got from AU Time, AU Result or AU Occupation Tools:
0 € < OGAP < 7.000 €: Ok.
7.000 € < OGAP < 10.000 €: Acceptable.
10.000 € < OGAP: Critical.
When it is critical, Users (either Operators or Experts) have to explain why a so high OGAP has been obtained. In
that purpose, they have to associate to the OGAP attributions defined on MANUFEX Program. There are four
levels of MANUFEX attributions. The first level is composed of:
Business: No demand
Extern: Independent of DSM (delay of the raw material supplier)
Maintenance: Due to equipments
Operation: Due to operations of production.
Then attributions from the following levels are used to go more in details. For instance, an OGAP could be
explained by:
Level 1: Maintenance
Level 2: Equipment Breakdown
Level 3: Vessel A
Once these attributions are defined, a graph OGAPs versus MANUFEX
Attributions can be plotted. In this way, performance killers are defined on the
production line. For instance, on Figure 37, efforts should be focused on
Maintenance until this Attribution looses its status of Performance Killer number
one. That can be done as well with Attributions Level 2 or 3 in order to find
equipments which meet problems on the line. This method is a real continuous
improvement one. Figure 37: TOP 10 Performance
Killers Level 1.
Thus, AU Tool is useful in order to determine main problems on the production line and even its bottlenecks. That
lets to oriented research area and Experts‟ studies in a way to obtain a higher productivity with costs as low as
possible.
3. Utilization and Configuration of OPN-AU Tool
OPN-AU Tool is a really powerful Tool which can allow real improvements in terms of productivity, control and
reactivity. In that sense, it has to be ease for use and fast to configure.
a. Utilization
In order to assure a fast detection of problems, the OPN-AU Tool main window represents the overview of all the
plant. The main screen is divided into various workshops representing each area and each workshop consists of
equipments, modelled by small boxes (rectangles), which belong to the area (Figure 38).
Basic knowledge
Initially, all boxes are white. As soon as a product/equipment combination is activated by opening a PO on MES,
they become colourful and the tracking of defined KOPs starts. (The colour code will be explained further in the
report). All Seclin plan can so be monitored in an easy way.
Configuration and Optimization of a Process Monitoring Tool 22 |
Chalmers University of Technology | The equipments‟ boxes have got many features and the most important one is by clicking on the box, the
monitored KOPs list appears. It lets Operators to get lots of information about the operation which is running
(Figure 39):
Product currently on the equipment is precised on the bottom.
Information about the operation such as batch Number, Product and activated PO /Phase, is available by
clicking on the yellow wheel.
A colour code is made to specify the kind of KOP monitored:
Brown: Process parameters (temperature, pH...) called Golden KOP – OPN Tool
Blue: Duration called Time KOP– AU Tool
Green: Start Delay called Start KOP– AU Tool
The current values of KOPs and current KOPs profiles‟ values (red and orange border areas) are
displayed:
MinMin = Low red border area
Min = Low orange border area
Valeur = Current KOP value
Max = High orange border area
MaxMax = High red border area
An access to the KOP curve is possible by clicking on “Detail”. It gives access as well to corrective
action trees if needed.
Reaction and correction
When a PO is activated on equipment, the corresponding box on OPN-AU Tool window becomes colourful by
respecting a special code:
Green: All KOPs are in their green area.
Orange: At least one KOP is in its orange area.
Red: At least one KOP is in its red area.
When an equipment box becomes orange or red, it is blinking and it is considered as an alarm. When
Operator notice an alarm, they have to follow a specific procedure:
1. Click on the concerned equipment‟s box. The KOPs list appears and the line of the wrong KOP is
blinking in orange/red (Figure 40).
2. Click on the “Detail” button of the KOP which has a critical state. Its curve, its values and its
corresponding corrective action trees appear (Figure 41).
3. Analyze the problem on the KOP curve.
4. Read the assigned corrective decision tree (Annexe III) which guides Operators to solve the problem.
5. Fix the problem.
6. Confirm the alarm.
Once an alarm is confirmed, the box does not blink anymore, but it keeps its orange/red value until the KOP
comes back to the green area.
Configuration and Optimization of a Process Monitoring Tool 23 |
Chalmers University of Technology | Two different ways are possible to confirm an alarm depending on its colour:
Orange alarms: Operators have to double click on the KOP line in the KOPs list.
Red alarms: Operators have to double click on the KOP line in the KOPs list and fill a confirmation
message to explain the alarm and the reaction he had to fix the problem. Moreover he has to assign
MANUFEX Attributions to this issue (Figure 42) if it is a KOP relevant for AU Tool.
When red alarms are confirmed, they are stored in an alarms listing and MANUFEX Attributions will be used in
the AU Tool (Figure 43). It sums up all information about alarms which happened during a given period (KOP,
value, Product, Equipment, Confirmation message). It is useful afterwards for pointing out problems which have
to be fixed for next batches.
Thus, OPN-AU Tool is really visual and its use results in an easy monitoring of significant parameters on the
production line. However, it is possible only under the condition of a good configuration of KOPs.
b. Configuration
Configuration is mainly performed on a part of the Tool called Product Builder. Only users defined as
Administrator can have access to it. Since KOPs in Seclin are product-equipment dependent, it is needed to define
KOPs for each product/equipment combination. That is why it is important Product Builder enables
Administrators to do this task in a convenient way.
Figure 44 represents Product Builder, divided into four different parts [14]:
Product Tree: It is composed of the list of products made in DSM Seclin. It has to be noticed that a same
enzyme, on two different steps of the production is considered as two different products (e.g. “Enzyme A
Inoculum” and “Enzyme A Fermentation” are not the same product in Product Builder). This area allows
assigning equipments and KOPs to products.
KOP definition Editor: It lets to define KOPs and their profiles.
List of all plant areas: It contains all areas of the Seclin plant. It was defined during the creation of the
OPN-AU Tool but it can still be modified via Plant area Builder in OPN Tool.
List of equipments assigned to the selected plant area: It shows equipments included in each plant area. It
can be modified via the MES Database related to equipments.
Product Tree
The products list automatically comes from a MES Database and is updated every six hours. To configure KOPs,
first, equipments where the product can go through, have to be assigned in Product Tree to the right Product.
To add a new equipment to a product, it has to be selected in its plant area and has to be
dragged and dropped from the equipment list to the product. Then the equipment appears
under the product.(Figure 45)
Figure 45: Equipment
assignment.
Once the equipment is added, its weight factor used in the Global Scoring
study is set in the Editor part (Figure 46).
Figure 46: Equipment Editor.
Configuration and Optimization of a Process Monitoring Tool 24 |
Chalmers University of Technology | KOPs which have to be tracked on this equipment can then be added by selecting assigned product
and equipment and pressing the “KOP” button in the Toolbar.
That creates a new KOP called Batch KOP1 under the equipment (Figure 47). Thus,
assignments on Product Tree are done and everything is ready to configure the KOP in KOP
Definition Editor.
Figure 47: KOP addition.
KOP Definition Editor
Figure 48 represents the KOP Definition Editor‟s Window. It allows defining completely a KOP, its conditions of
tracking, its impact on the Scoring and profiles it should fit during the operation. [14]
Definition of the KOP name (Temperature, pH...)
Fields reserved for further description (Autolyse Temperature...)
Process Order and Phase (From MES)
KOP will be tracked only within the selected Process Order/Phase.
Condition which has to be filled to start the tracking.
E.g.[Temperature tag]>N implies that the tracking begins only once temperature is higher than N°C.
It is an important feature called: First Condition for convenience. If this condition is not filled, KOP
is tracked but KOPs profiles are not activated so no alarm can rise.
It is described more in detail below.
Tag or Formula that is tracked by the KOP. In a certain way, it can be used to set a Second
Condition:
E.g. IF ([temperature tag]>N1; [pH tag];N2) implies when temperature is higher than N1°C, pH is
tracked otherwise, the KOP value is set to N2, usually equal to the optimum. That lets to avoid
aberrant alarms if it is known that at lower temperature, pH cannot be controlled.
Definition of profiles that the KOP should fit. In real terms, it defines green, orange and red areas as
time functions:
Definition of the corrective action trees.
Fields from part 6 on Figure 48 are really important regarding the Scoring and conditions needed to rise an alarm.
Indeed, some conditions can be set to avoid aberrant alarms. Thus many features are included in this part of the
Editor:
Batch KOP Type: Administrator precises the KOP type implemented, meaning a process parameter (Golden
KOP), duration (Time KOP) or a start delay (Start KOP).
Alarm Delay: It specifies the time span the following alarms will be suppressed after an alarm has been
raised the first time. For instance, if Alarm Delay is set to 10 minutes and an alarm raises for this KOP, all
following alarms within the next 10 minutes are suppressed.
Activation alarm: It is a delay for raising an alarm when a KOP is in orange or red areas. For instance, if it
is set to 10 minutes, a red alarm raises only when KOP values are in the red area longer than 10 minutes.
That is useful to avoid aberrant alarms such as when a peak value is obtained because of a malfunction of
a sensor.
Configuration and Optimization of a Process Monitoring Tool 25 |
Chalmers University of Technology | Coefficient: (i.e. Weight Factor) Administrator specifies the weight factor of the KOP for the Scoring study.
Delay Démarrage: (i.e. Start Delay) It is a delay for rising an alarm after the beginning of the tracking. It
mainly lets to avoid alarms during fillings of vessels where KOPs‟ values are not really relevant.
Suivi Moyenne (i.e. Average) It lets to monitor an average of KOP values instead of tracking simple values.
It is used to minimize possible small fluctuations of parameters.
Zone de Latence (i.e. Deathband) It specifies a deadband where alarms still raises even if KOPs came back
in green area. It is not used for the moment on Seclin plant.
The rest of configuration depends on the Batch KOP Type selected by Administrator:
Time KOP
Start KOP
Golden KOP
Time KOP
This KOP Type enables to track duration of a certain PO and/or Phase. Administrator just needs to select the
PO/Phase he wants to track and enter the duration alarm borders in minutes. For that he defines the High and
HighHigh limits defined in collaboration with Experts. He can specify as well for which Tool this monitoring is
relevant (Figure 49):
For OPN: Alarm appears only in the OPN deviation list and in the
Scoring.
For AU: Alarm appears only in the Asset Utilization deviation list so
it should be found in the AU database and a deviation cause can be
assigned to.
Figure 49: Configuration of a Time KOP.
Start KOP
This KOP Type enables to compare the actual start time of a phase with the planned start time stored in the MES
system. An alarm rises if the delay of the operation is higher than the „High‟ and „HighHigh‟ limits initially
defined. Moreover, like in the Time KOP, Administrator has to specify for which Tool the KOP is relevant.
Golden KOP
This KOP Type enables to track Golden batch curves which mean process parameters. The tracking can be
performed either on a single parameter or a formula of different parameters.
As explained before, process parameters are expressed by Tags created by Logica, an Austrian supplier which can
make Tags available in MES on order from DSM:
They can be concrete parameters from ABB such as pH values got from a sensor on a vessel.
They can be a virtual one such as a Tag defined as a parameter which is equal to 0 when a valve is closed
and 1 when it is open.
Administrator has to select the PO/Phase he wants to track and drag the parameter from a tagbrowser into the field
IP21 tagname (number 7 on Figure 48).
Configuration and Optimization of a Process Monitoring Tool 26 |
Chalmers University of Technology | Two other significant fields have to be filled:
Definition of Batch profiles: It is done by clicking on ”Define profile” button (number 8 on Figure 48).
Administrator is then navigated to the profile definition window (Figure 50). In real terms, he defines
values of green, orange and red border areas versus time. It is possible to import profiles from Excel. It
allows a real saving of time because curves can be studied on Excel and then exported directly to OPN-
AU Tool. Some rules have to be respected during this setting:
At least two lines for each profile have to be defined (a starting line at hour 0 and a end line of
the profile)
Of course, values should be set such as LowLow ≤ Low ≤ Optimum ≤ High≤ HighHigh.
Definition of corrective action trees: It is done by clicking on ”Decision Tree” button from KOP
definition Editor (number 9 on Figure 48). Actually they are downloaded from Excel files previously
made by Administrator.(Figure 51). So it lets to add a link to the Tree stored on the network.
Thus, Administrator is able to define quite precisely what Production Staff want to monitor and the way that they
want alarms to be raised. It allows to have an efficient Tool for controlling the production. However, a perfect
configuration is needed to assure a good monitoring and benefits from it.
Configuration and Optimization of a Process Monitoring Tool 27 |
Chalmers University of Technology | MASTER’S THESIS’ STUDY
With OPN-AU Tool, DSM Seclin has a powerful tool, unique in the group, to monitor its production line. It is
now a reference on a worldly scale. It has recently even been awarded within the group by a DSM Award for
Excellence (DFS Exceptional Achievement Award) in June 2009. However; some improvements are still needed.
Indeed, because of the diversity of products made on the plant, its configuration is not finished yet and Operators,
Experts and SPS members have noticed some problems in what is already done. That resulted in many aberrant
alarms which have generated a loss of reliability regarding the Tool. So Operators were less careful to alarms
thinking that they were mainly not relevant. In consequence, the company decided to hire a student in order to
increase the OPN-AU‟s efficiency and so improve the process monitoring.
A. Mission Description
1. Initial State
A batch production implies process conditions depending on products. KOPs are so product and equipment
dependent which results in a need of an accurate configuration of OPN-AU Tool. But it is quite time consuming
and quite difficult to adapt monitoring to each product and each operation.
These difficulties have been pointed out by the noticed delay of the company in configuration of their new
products and the absence of improvement on what is already implemented. Indeed, at the start of this study, OPN-
AU Tool presented two main problems:
A lack of configuration: Several new products‟ productions cannot be monitored because they are not
configured on the Tool.
A maladjusted configuration: Lots of aberrant alarms have been notified by Operators during the
production. For instance, pH sensor is not covered yet by the liquid in a vessel but PO is already
activated; so the tracking begins and the no-relevant pH value generates an alarm. In addition, some
KOPs profiles do not fit with what it is really going on.
On January 2010, almost 40% of red alarms were considered as aberrant.
Thus, even if OPN-AU Tool has been recognized as a real powerful Tool, continuous improvements have to be
performed to guarantee a real efficient monitoring of the production line (OPN) and relevant studies regarding the
productivity (AU).
2. Aimed Improvements
This Master‟s Thesis aims to improve the efficiency of OPN-AU Tool in various fields:
Configuration of new products
Optimization of initial configurations
Project Management
a. OPN-AU Tool: Configuration of new products
In three years, in collaboration with its headquarters, DSM Seclin found new strains and so new Enzymes to
produce. These innovative productions have to be monitored as well for each step of the production line. In
consequence, this Master‟s Thesis includes configuration of new products on all the plant.
Configuration and Optimization of a Process Monitoring Tool 28 |
Chalmers University of Technology | b. OPN-AU Tool: Optimization of initial configurations
Being a recent Tool, OPN-AU Tool has been configured quite quickly in order to cover as much operations as
possible. This implementation has resulted in definition of simplified KOPs profiles in certain areas especially in
DSP and Formulation. Indeed, Fermentation has been configured for almost all products in a really relevant way;
however, KOPs profiles of some other steps have been reduced to the final desired values of the KOPs. No
evolution of parameters in time has been implemented. In consequence, many alarms used to rise during
operations because KOPs profiles are not adapted to the evolution of parameters. Even if some modifications are
needed in all areas, this Master‟s Thesis mainly focused on optimization of the Extraction and Formulation
monitoring which includes more than 250 products and 45 equipments.
c. Shift Report Tool: Project Management
Shift Report Tool, one of the future projects of DSM Seclin, should be included in OPN-AU Tool. It will let
Operators to do their shift reports on computers whereas currently they do it on paper sheets. It will so allow
Users to save time thanks to its auto-filled fields from MES, ABB and OPN. It will also let to avoid mistakes,
oblivion, and so to assure a better communication between teams.
Shift Report Tool has been imagined by DSM Seclin but as OPN-AU Tool, it will be made by ISG. A User
Requirement Specification (URS) was already made and ISG gave a first offer regarding the order. As part of this
Master‟s Thesis, the price has to be discussed and then project management has to be performed. It especially
includes a participation in the writing of the Functional Design Specification (FDS), a validation of the Tool and
the training of Staff once it will be implemented on the plant. It validation will be performed through different
ways. First a Factory Acceptance Test (FAT) will be done at the Supplier place to do some tests on the Tool.
Then, a Site Acceptance Test (SAT) will let to test the Tool directly on the plant. Finally a continuous
optimization will be assured during six months in order to guarantee its efficiency.
Thus, concerning both production and research areas and giving opportunities to practise project management,
this Master‟s Thesis represented an overall of what can be performed by an engineer. It implied a really good
knowledge of processes performed on the plant and also active participation of Production Staff.
3. Available Means
Because of their plurality, targets of the Master‟s Thesis force to know almost everything about DSM Seclin
Plant. Several informatics Tools and experience of Production Staff have been really helpful to go in that sense
and to assure a good understanding of what is going on the production line.
The first three weeks of this Study began with a learning of the production procedures, on each workshop of the
Seclin plant. That was indispensable in order to configure new products all along the production line for instance.
It has been performed by following Operators during their work period and observing precisely how they do their
job. This training was completed by meetings with all Experts who could explain, more on a scientific point of
view, the different aspects of operations.
Once, the production line, the operation conditions, the management system and the way to configure the OPN-
AU Tool were understood, the real work could begin. It has been done in collaboration with the SPS members and
Experts and thanks to the use of different informatics Tools. Indeed, many means were available.
Configuration and Optimization of a Process Monitoring Tool 29 |
Chalmers University of Technology | a. OPN-AU Tool: Configuration of new products
Configuration of a new product implies the study of each step of the production plant. It was mainly done with
SPS members. Each one is responsible of specific products. During meetings it was decided which KOPs should
be monitored on each operation and which profiles they should fit. The main difficulty on this case is the absence
of knowledge. By definition, because they are new products, no data is available to rely on and to take right
decisions. In that sense, after a first configuration, some changes could be needed after having analyzed different
batches.
b. OPN-AU Tool: Optimization of initial configurations
By including various measures, optimization of initial configuration has been performed on different ways. It
regroups three main actions:
Continuous configuration adaptation.
Continuous issues‟ solving.
Complete modification of maladjusted configurations.
Continuous configuration adaptation
After each batch production, SPS members and Experts have to analyze results, yield and KOP curves obtained
during the operations. In that purpose, they calculate different OGAPs on AU, define different Scorings on OPN,
and study some correlations between KOPs choice and obtained yield. They can deduce from that if the KOPs
listings are relevant. Moreover, by observing KOPs curves, they can notice if initial KOPs profiles are adapted.
According to their results they planned a meeting in order to talk about all possible improvements in Product
Builder.
Continuous issues’ solving
Each morning, production Experts, maintenance Leader, Planning Builders and the production Manager have a
meeting to talk about problems which happened the previous day in the plant. These issues are described in a
Database called “Deviation” (Figure 52). As soon as a problem occurs, Operators write a deviation describing it,
its causes, its consequences and how they tried to fix it. Each deviation is then commented at the meeting and
assigned to a manager who is responsible of the problem‟s solving.
When OPN-AU Tool‟s Users notice problems regarding the Tool such as a maladjusted configuration, deviation
is treated by analyzing the issue, comparing with procedures and making needed changes in configuration. An
example of this kind of measure will be presented further in the report.
In the same way, several optimizations of the Tool have been performed by consulting the KOP alarms‟ listing
(Figure 43). Operators have to write a confirmation message in order to validate red alarms. Through this
procedure, they sometimes point out problems of configuration so it can be a good source of information for
modifying some maladjusted configurations. It was as well completed by meeting each operator every morning
before the Deviation meeting to talk with them about problems they might have noticed or alarms which were
blinking.
Complete modification of maladjusted configurations
As said before, OPN-AU Tool is quite recent and to assure an operational Tool, some of its configurations have
been done rapidly, simplifying KOPs profiles which are now source of aberrant alarms. That includes mainly two
operations:
In Extraction: The killing/harvesting phase of the product before sending it to membrane filters press.
In Formulation: The formulation phase, regrouping the preparation of the product to get the final one and
its last filtration through the second plate filters (FAP2).
Configuration and Optimization of a Process Monitoring Tool 30 |
Chalmers University of Technology | That is why the Master‟s Thesis has been mainly focused on these two optimizations. Added to the Operators‟
experience, two informatics Tools were really useful to succeed in implementing improved KOPs:
AspenBatch 21DetailDisplay (IP21): Extracting data from MES (relying on POs), it lets to find, on which
equipments and when a given batch production took place (Figure 53).
Aspen Process Explorer: It plots values of each parameter assigned to a Tag versus time (Figure 54).
Using first IP21, it can be defined on where (which equipment) and when (date and time) an operation took place.
Then, on Aspen Process Explorer can be plotted and so analyzed curves of different parameters of a specific
operation.
For instance, the temperature‟s trend of an Enzyme A on the fermentation phase could be studied if at least one
batch number of this operation is known: On IP21, it is deduced that the operation took place in Fermentor 1 on
January the 1rst 2010 at 8:00 am.
Then it is easy to obtain the temperature‟s evolution of the operation by going on Aspen Process Explorer and
plotting the Tag corresponding to the temperature sensor on Fermentor 1 at the relevant time. It is a good way to
see why current configurations do not fit the reality and how it could be improved. Some examples are going to be
detailed in the next part of the report.
Thus, this program includes many different aspects but the availability of many Tools lets to configure many
parameters in a restrictive time.
B. Performed Improvements
This part presents some examples of performed configurations and kinds of problems which it had to face up to.
The ones described here illustrate main improvements implemented during the Master‟s Thesis.
1. Example of an Extraction’s configuration
Once fermentation is over, cultivated strains are killed (killing phase) and the product is prepared to go through
membranes filter press (harvesting phase). These steps are usually performed under the PO Killing/Harvesting
(PO K/H) which consists of two phases: the killing phase and the harvesting phase.
It has to be noticed that not all POs have a killing phase and for some materials it is not needed to kill the broth
and POs only have a harvesting phase.
When a withdrawal is done from a fermentor, PO should already be selected by Operators in MES interface on
the vessel where the withdrawal will be transported to. They should define in ABB the right destination vessel
before opening the bottom valve on the fermentor to transfer the broth. So PO K/H is activated and the Killing
phase is automatically set when a killing is needed (Figure 55). The Harvesting phase is then selected when for a
first time operators dosed a specific adjuvant A into the vessel.
Figure 55: Description of PO Killing/Harvesting.
Configuration and Optimization of a Process Monitoring Tool 31 |
Chalmers University of Technology | The existing configuration on OPN Tool was not adjusted to these operations in the sense that all KOPs profiles
were equal to the final values aimed at the end of the harvesting phase whereas a Killing phase for instance
implies a willingly increase of temperature (Figure 56). As a consequence, many aberrant alarms used to rise.
That is why the configuration had to be modified.
After studying accurately all the process and Operators‟ actions, two main problems have been remarked:
Withdrawals.
Transition between killing and harvesting phases.
Withdrawals
Usually, during fermentation, withdrawals are sent to a vessel where they are cooled waiting for the next ones to
perform the Killing. Thus, the cooling has been first configured with KOP profiles which decrease versus time.
However, it has been noticed the arrival of the hot next withdrawals into the vessel used to increase temperature
in such way that an alarm raised (Figure 57). The problem is that nobody knows exactly when transfers occur, so
KOP profiles could not be easily configured.
After long talks with Operators, the first idea was to put a condition on the fermentor‟s bottom valve such as no
alarm raised when the valve is open for the transfer of a next withdrawal. It was not a good way of
implementation since after closing the valve, the mixing between withdrawals is still hotter than the KOPs
profiles configured for the cooling. So alarms were not totally avoided.
A second idea was to put a condition on the vessel‟s weight in the “Second Condition” field of the KOPs Editor.
Indeed, a second withdrawal implies an increase of the weight. So it has been tried to configure operations by
creating one Temperature KOP per withdrawal. All KOPs would be tracked during whole the operation but they
should be simultaneously relevant. For instance, let imagine a vessel with two withdrawals. KOPs should be the
ones below:
Temperature ST1: It is relevant when only one withdrawal is in the vessel. It is used to control its cooling.
Its configuration implies two main conditions:
First condition: [Weight] > x.
It implies a tracking when the weight is higher than x tons to avoid alarms due to a not-enough
filled vessel.
Second condition: IF ([weight tag] < x’; [temperature tag]; N).
It implies KOP equal to the temperature value when the weight is lower than x‟ tons. When it is
not, KOP is set to N, which is equal to the Optimum. So, it avoids aberrant alarms and
temperature is so monitored on Temperature ST2.
Temperature ST2: It is relevant when two withdrawals are in the vessel. It is used to monitor the cooling
of the new mixing. In the same way, its configuration implies two main conditions:
First condition: [Weight] > x.
It implies a tracking only when the weight is higher than x tons to avoid alarms due to a not
enough filled vessel.
Second condition: IF ([weight tag] > x’; [temperature tag]; N)
It implies a KOP equal to the temperature value when the weight is higher than x‟ tons and a
KOP set to N, the optimum, when it is not. So that could avoid an alarm before the second
withdrawal arrives.
This configuration is time consuming especially when fermentation implies many withdrawals. Moreover, it has
been find out that it is not a good way to monitor temperature in the sense that the weight of a withdrawal is not
constant for batches of a same product. So in the Second condition, it is hard to determine the correct value of x‟
which should define which profiles have to be followed. So this solution has not been accepted.
Configuration and Optimization of a Process Monitoring Tool 32 |
Chalmers University of Technology | Finally, after various other tests, two ways have been kept depending on the product characteristic to fix the issue.
The first one consists in implementing KOP profiles with larger OWs in order to include the increase of
temperature due to withdrawals (Figure 58).
The second one keeps initial KOPs profiles, taking into account only the first withdrawal‟s cooling
(Figure 59). However an activation alarm, which represents a delay for raising an alarm when a KOP is in
orange or red areas, is added and set equal to the theoretical time needed to cool the next withdrawal. That
implies, on the curves point of view, a KOP outside the red borders but no alarm rises on the vessel so
operators are not careful to the drift.
Transition between killing and harvesting phases
The Killing phase generally implies to keep the product under specific conditions during a given period (high
temperature and a low/high pH). After that, it is cooled and its pH is modified to put the product under its
harvesting conditions.
In that case the problem is mainly due to the batch production. Products made in the same time than another one
are rarely the same for different batches. It implies variability on the operators‟ occupation. So after having cooled
the mixing, it is difficult to know in advance when exactly the killing begins because it depends on the availability
of Operators.
After several tests, two different solutions have been found out depending on the way used for the killing:
To create two KOPs for a same parameter: One for the cooling and one for the killing. In that case,
different “Second conditions” have been set to track the killing such as following temperature Tag as
soon as Operators set on ABB a pH setpoint equal to N.
To analyse old batches and define an average duration of each operation. So parameters are monitored
through only one KOP and both cooling and killing are tracked with the same KOPs profiles (Figure 60).
These configurations of the extraction phase mainly rely on curves of the old batches studied on Aspen Process
Explorer. They have been made for more than 40 products and 10 vessels which result in 400 combinations. After
some checks directly performed on the plant, it seems that results are pretty conclusive even if some optimizations
still need to be done (Figure 60).
2. Example of a Formulation’s configuration
Formulation is the operation where the product is prepared in order to fit exactly with the final product ordered by
customers. Made under PO Liquid Formulation, this operation is composed of two phases in addition to the CIP
(Figure 61):
Preparation phase
Filtration phase
Figure 61: Description of the PO Liquid Formulation.
Configuration and Optimization of a Process Monitoring Tool 33 |
Chalmers University of Technology | The filtration phase, performed by a Plate Filter (FAP2), is optional and for some products the sequence of
preparation and filtration can happen twice or more, it depends on the recipe. When filtration takes place, the
destination is a vessel (Figure 62).
Figure 62: Filtration description.
Initially, the configuration of Formulation was maladjusted. Indeed, during PO Liquid Formulation, KOPs
profiles were set equal to values aimed at the end of filtration even if the Preparation phase is performed under
different conditions than the Filtration one (Figure 63). That is why all configurations had to be readjusted.
The main problem is that a same vessel can be used for both Preparation and Filtration phases. Indeed, during the
filtration, Filtration phase is activated on two vessels; the destination one and the source one. So if a KOP is
defined under the PO Liquid Formulation/Filtration Phase, the tracking was performed without checking if it is
the source vessel or the destination one whereas assigned KOPS profiles should be different. Thus a Second
condition had to be found in order to differentiate KOPs profiles according to the vessel‟s feature. For instance,
referring to Figure 62, it has to be specified if KOP has to follow the KOPs profiles A or the KOPs profiles B.
Different configurations have been tested after having studying in detail the operators‟ action. Nothing seemed to
be relevant. In consequence, a creation of new virtual Tags has been ordered to ISG. Called ETAP, they should
specify which operation is taking place into a vessel. One Tag per vessel has been created and its value can vary
from 0 to 4:
ETAPVXX = 0 when no PO is open and so nothing occurs into the vessel
ETAPVXX = 1 when the preparation phase is performed.
ETAPVXX = 2 when the Filtration phase is open and the vessel is used as source of FAP2.
ETAPVXX = 3 when the Filtration phase is open and the vessel is used as destination vessel of FAP2.
ETAPVXX = 4 when the Filtration phase is over but the vessel still store the product.
Now that a vessel could be defined, many KOPs have been defined under a same product and for a same vessel.
For instance, even if just temperature is monitored 3 different KOPs are configured:
Temperature Preparation: KOPs profiles correspond to the ones adapted to the Preparation phase. No
specific condition is set because the Preparation phase is selected in KOP Editor.
Temperature Filtration: KOPs profiles correspond to the ones needed when the vessel is emptying to
transfer the product to FAP2. It is performed under the Filtration phase. To specify that it is the source
vessel, a Second Condition is set: IF ([ETAPVXX] = 2; [Temperature Tag]; N). It means KOP is equal
to actual temperature when ETAPVXX=2 and otherwise, it is set to N, the optimum which does not
trigger any alarm.
Temperature Reception: KOPs profiles correspond to the ones needed when the vessel is filling by FAP2.
It is performed under the Filtration phase. To specify that it is the destination vessel, a Second Condition
is set: IF ([ETAPVXX] = 3; [temperature Tag]; N). It means KOP is equal to actual temperature when
ETAPVXX=3 and otherwise, it is set to N, the optimum which does not trigger any alarm.
This configuration has been mainly performed by analysing curves of old batches on Aspen Process Explorer
(Figure 64). It included more than 150 products and 19 different vessels and so around 2.800 combinations.
Unfortunately, efficiency of this new configuration has not been checked yet but it seems to be the best alternative
for the moment to monitor as much as possible KOPs and avoid aberrant alarms. It can monitor each step of
Formulation.
Configuration and Optimization of a Process Monitoring Tool 34 |
Chalmers University of Technology | 3. Example of a Deviation’s solving
One big problem which impacted on each equipment and each product, has been pointed out: The wrong KOPs‟
values got from a sensor which is not covered by the product because of an not-enough filled vessel used to
activate alarms. In consequence, during each vessel‟s filling and emptying, alarms used to raise (Figure 65).
This issue hs been notified by many Operators, directly or through deviations. (Figure 66). In consequence, a
condition of tracking avoiding that has been looked for:
A research of PID plans and digital control design has been made through the Engineering project department„s
archives. The purpose was to define position of sensors on each vessel and all dimensions of equipments
(diameters, height and thickness of the double layer...).
Only few data have been find out so all needed measurements have been finally taken directly on the plant in
collaboration with the Mechanical department.
Once all parameters were collected, a minimal volume of product assuring that sensors dip into the liquid has
been defined for each vessel.
However, nothing can monitor the volume of product present in vessels. Only a condition on the vessel‟s weight
could be done. So samples of different products on different steps of the production have been collected and
density of liquids has been defined in a laboratory. Thus the minimal volume needed is translated in a minimal
weight.
For each vessel the minimal weight of liquid needed to assure that sensors are covered by the product before
begining the KOPs tracking has been defined. In real terms regarding the OPN-AU Tool, for each KOP, in the
Editor field), has been applied a First Condition on the weight as followed (number 5 of Figure 48): [tag of the
vessel’s weight] > X Tons.
Including more than 60 vessels and almost 350 products, this implementation decreases considerably the number
of aberrant alarms by deleting ones due to vessel‟fillings and emptyings. However it has to be noticed that this
condition uses the “First Condition” in the KOPs Editor for every configured KOPs. In consequence, it highly
limits the field of action to fix other issues since only one can be used per KOP.
All these examples do not represent all optimizations and configurations performed on the OPN-AU Tool because
too many different changes have been made. However, they illustrate the way of thinking to fix issues and to
configure specific operations. The result of all the work performed during this Master‟s Thesis is present in the
following part.
C. Results and Communication
This Master‟s Thesis aimed to improve and develop the OPN-AU Tool in different ways. Since the Tool is used
by Operators directly on the plant, impacts of performed modifications could be easily noticed and
communication sessions had to be scheduled in order to explain to users changes performed on the Tool during
the Master‟s Thesis‟ period. The main purpose was to rekindle enthusiasm, active participation and trust of
Operators in OPN-AU Tool.
Configuration and Optimization of a Process Monitoring Tool 35 |
Chalmers University of Technology | 1. Results on OPN-AU Tool
a. AU Tool Improvments
On AU Tool, two main implementations have been performed:
All borders defined for Start KOPs and Time KOPs in Fermentation have been updated in order to be
adapted to changes on the process line decided by SPS members such as a longer Fermentation.
New Start KOPs and Time KOPs have been implemented in DSP, mainly on the Ultrafiltration and
Chromatography steps in order to let Experts to analyze delays noticed during such operations.
b. OPN Tool Improvments
Too many different changes have been performed on OPN Tool to be able to list done improvements. However, it
can be noticed that:
Around 50 deviations have been assigned to OPN Tool and more than 75% of them have been fixed at the
writing of the report.
Three new products have been totally implemented on each step of the production line.
Configurations of Extraction and Formulation Phases have been totally readjusted which include more
than 200 products and 45 equipments and so around 20.000 configurations of KOPs.
Impacts of these implementations can be analyzed through different calculations:
Percentage of aberrant alarms: From the beginning of the Master‟s Thesis, twice a day percentage of the
alarms which are aberrant have been calculated. Its trend along months illustrates impacts of new
implementations (Figure 67). It can be seen a high decrease of the number of aberrant alarms from almost
40% to less than 20% in four months. That lets to conclude that performed implementations have real
benefits on OPN Tool‟s efficiency. However, the Tool still needs improvements.
Evolution of the Scoring got on the Extraction area: Many configurations have been adjusted in the
Extraction area. Evolution of its average scoring got on operations made per week can illustrate efficiency
of KOPs profiles‟ definition. Depending on alarms, the Scoring translates problems got during operations
but it lets to see as well if KOPs profiles are adapted to the process. On Figure 68, it can be seen a real
improvement versus weeks on this step which lets to conclude that performed configurations are quite
good.
Evolution of the Global Scoring: In the same way, evolution of the Global Scoring got on DSM plant
versus week can translate the efficiency of implementations. A global improvement is noticed with an
increase of the Global Scoring from 76.6 % to 92.9 % in less than four months (Figure 69). On Figure 69,
some Scoring values did not include results got on drying phase because of a problem which happened on
MES and obstructed the monitoring.
Thus, for both OPN and AU Tools, results are really positive and imply a higher efficiency of the OPN-AU Tool
and so a better process control. However, improvements are still possible and configuration of Formulation still
has to be analyzed.
2. Results on the Shift Report Tool
Regarding Shift Report Tool, first, by talking and negotiating with ISG, the price of its implementation has been
reduced from 48,000€ to 26,000€. In consequence, the offer has been presented to the Site Director, the Leader
Expert and the Production Manager in order to have their agreement to perform it.
Configuration and Optimization of a Process Monitoring Tool 36 |
Chalmers University of Technology | Shift Report Tool has been finally ordered and its Functional Design Specification has been written in
collaboration with ISG. It contains description of each feature and each characteristic of the Tool. It has to be
done in a careful way because the Austrian company relies on it to create the Tool. Then it has been validated by
several managers of the plant.
Its creation is runaway and the Tool should be finished at the end of June. Once it will be completely performed,
the validation period will begin in collaboration with ISG. It will be followed by its implementation on DSM
Seclin plant. Finally a training period during around three weeks will be needed to explain to Operators how to
use it.
3. Communication
Many modifications have been performed on OPN-AU Tool. Communication to Operators was so indispensable
to assure a good use of a Tool and increase their motivations, their belonging to the Tool and so their active
participation in its improvement. It implies two main measures:
Reminding
Motivation and Information
a. Remindings
Because of the initial high number of aberrant alarms, operators did not really take care anymore to alarms. It was
illustrated by the absence of alarms‟ confirmation and so low Scorings. After having performed improvements, it
was necessary to let operators claim the Tool again and remind them what they have to do regarding it.
Two examples can be explained:
Too many times Operators did not confirm alarms and did not inform about problems noticed on the
Tool‟s configuration. Thus information campaign was indispensable to remind them how they have to use
the Tool and they have to react when an alarm rises. It has been made by speaking with them and by
writing a procedure published in each control room of the plant (Annexe IV).
At the beginning of their work period, Operators check pH of vessels by collecting a sample and defining
its pH value in laboratory. If the result is different than pH indicated by the vessel‟s sensors, they have to
adjust sensors manually. Unfortunately they often do not do it. It results in wrong pH values got from
sensors and so wrong Tag value. It implies of course wrong KOP values and so aberrant alarms. That is
why, it was needed to explain and/or remind operators that it is important to do it. It has been done by
speaking with them and writing a small procedure available in each DSP control room (Annexe V).
b. Motivation and Information
In order to assure a right utilization of OPN-AU Tool, it was important to motivate Operators. It has been mainly
performed by showing them the Tool‟s efficiency, rewarding them and informing them about new
implementations.
In that sense the Scoring got on each operation performed the previous week was calculated every Mondays. Thus
Global Scoring and the average Scorings got on Fermentation, Extraction and Drying of the previous week was
deduced. These results have been then published in each control room. Figures are completed by some remarks in
order to reward Operators, specify where there was a problem and analyze trends. (Annexe VI).
Moreover, an informative session is planned for the end of the Master‟s Thesis to describe improvements and new
configurations performed on their Tool, OPN-AU Tool. It will be the occasion as well to introduce in few words
Shift Report Tool and remind them some features of OPN Tool that they do not handle with. (Annexe VII).
Configuration and Optimization of a Process Monitoring Tool 37 |
Chalmers University of Technology | CONCLUSIONS
Aiming to the improvement of OPN-AU Tool, the Master‟s Thesis has met the objectives set.
After having collected the needed information, many problems on initial configurations have been solved, the new
products‟ monitoring have been implemented, the Staff training has been improved and a following up of
implementation of the performed changes has been performed. Thus, the Tool which monitors the entire
production line is more efficient and is generally better used by the production staff. In addition, Shift Report
Tool project has seen a good progress since it is currently ordered and in state of realization.
However improvements are still needed on OPN-AU Tool and Shift Report Tool still has to be implemented on
the site. Furthermore, the process control depends not only on the Tools. It must be seen in a wider environment
and especially on the willingness on production Staff, to move things in the right direction. A tool is effective
only if one makes proper use. This should therefore be included in a global trend where the human dimension is
very important.
From a personal standpoint, this internship allowed me to discover new facets of the engineering profession as
designing tools, training of Staff and some managerial aspects related to the implementation of new practices
within Production Staff. It gave me the opportunities to work on many process equipments and so increase my
knowledge in process engineering.
This mission has also helped me to develop my sense of teamwork by operating in various sectors, to let myself to
use a powerful tool and also to improve my project management through Shift Report Tool.
The only problem was actually writing this report because it is very difficult to describe the construction and
operation of a computer application. The best solution remains a real use.
Thus, this Master‟s Thesis is a complete success in terms of training and performance. It is rare in an internship to
be given such responsibilities and such freedom of action and decision.
To conclude, optimization phase of OPN-AU Tool is on tracking. However, we must continue these efforts by
showing the importance and usefulness of the tool. The interest here is not to lose the general rekindled
enthusiasm.
Prospects are now to finish implementation of Shift Report Tool and perhaps consider other applications which
may be included in OPN-AU Tool such as the monitoring of actual breaks observed on operations.
Configuration and Optimization of a Process Monitoring Tool 39 |
Chalmers University of Technology | NOTATIONS
3/8: Organization of work where 3 shifts work 8 IP21: AspenBatch21 Detail Display.
hours each to cover 24 hours a day.
ISG: Austrian supplier of DSM.
ABB: Monitoring and digital control automaton use
KOP: Key Operating Parameter – Important
on the plant.
parameter which has to be monitored on
AU: Asset Utilization - Work Process of MANUFEX production.
to study productivity‟s performance.
KPI: Key Performance Indicator.
BU: Billion of Units (Enzyme Unit).
KPP: Key Process Parameter.
CIP: Cleaning In Place.
Logica: Austrian Supplier of DSM.
CODIR: Management Committee (Comité de
MANUFEX: MANUfacturing Excellence.
Direction).
CP: Counter Pressure.
MES: Manufacturing Execution System.
MPC: Maximum Proved Opportunity.
Déviations: Database to collect incidents or
problems noticed on the plant. OGAP: Opportunity GAP.
DFS: DSM Food Specialties. OPN: Operate Plant Normal – Work Process of
MANUFEX to perform the process
DHW: Deca Hectoliter Week.
control.
DOC: Measurement of the dissolved oxygen.
Orange area: Area of OW where process knows a
DSP: Down Stream Processing- It includes notable drift.
extraction, Filtration, concentration and
OUR: Oxygen Uptake Rate.
Formulation Phases.
OW: Operational Windows – Curves of KOPs
DSM: Dutch States Mines.
Profiles to define green orange and red
Enzyme: Active protein which makes it possible to areas.
transform a complex natural substance into
PO: Process Order.
simpler and more easily absorbable substance
PO K/H: Process Order Killing/Harvesting.
Expert: Technical and Process Responsible of a
production sector. Red area: Area of OW where process knows a
critical drift.
FAM: Membranes Filters press.
SAT: Site Acceptance Test.
FAP: Plates Filter.
SPS: Scientific Production Staff.
FAT: Factory Acceptance Test.
URS: User Requirement Specification.
FDS: Functional Design Specification.
Work Process: Unit of MANUFEX program.
Green area: Area of OW where process is under
control.
Configuration and Optimization of a Process Monitoring Tool 40 |
Chalmers University of Technology | REFERENCES
[1]: DSM Business Group – Our Strategy – Presentation – DSM – 2008.
[2]: DSM Food Specialties Seclin – Welcome to DSM Seclin – DSM – 2008.
[3]: DSM Food Specialties – Usine productive - Formation 2001 – DSM – 2001.
[4]: Fabian Flocard – Mise en place d‟Operate Plant Normal – Rapport de stage – Polytech‟Lille – 2007.
[5]: Marie-Pierre Guinchard – Notions de Biologie et Microbiologie – ENSGTI – 2007.
[6]: Cédrik Lallard – La sécurité est mon amie, il faut l‟aimer aussi – PLON – 2008.
[7]: Jean-Jacques Bimbenet, Albert Duquenoy and Gilles Trystram – Génie des procédés alimentaires -
Des bases aux applications – DUNOD – 2002.
[8]: Seaders J.D and Henley E.S. – Separation Process Principles – 2nd Edition – Wiley – 2006.
[9]: Anders Rasmuson – Lectures of Advanced Chemical Engineering and Process Analytical Technology
– KBT110 – Chalmers University of Technology – 2009.
[10]: Mathieu Mory – Séparation Mécanique – Cours de 2ème année – ENSGTI – 2007.
[11]: DFS Sunrise – iMES - Manuel Utilisateur – DSM – 2007.
[12]: DSM Sunrise – DFS Sunrise Seclin Project - MES - SAP Integration – Version 0.01 – DSM – 2007.
[13]: Fabian Flocard and Xavier Driesen – Operate Plant Normal - Asset Utilization - Manuel de
Formation – Programme MANUFEX – DSM – 2008.
[14]: ISG IT & Automation – OPN-AU Product Builder – ISG. – 2007.
[15]: Fabian Flocard – Bilan Test Correlation – DSM – 2008.
Configuration and Optimization of a Process Monitoring Tool 41 |
Chalmers University of Technology | Development of a decision support tool for operational optimization of the steam utility
system at Preemraff Lysekil
SEKPOOM KOBJAROENKUN
JOHAN GUNNARSSON
Department of Space, Earth and Environment
Chalmers University of Technology
Abstract
Steam is of high importance for an oil refinery. It is used as a heating media and to
generate mechanical shaft work. Steam is produced by heat recovery units and steam
boilers. The steam boilers are fuelled mainly by internally produced combustible gas
(refinery gas). When there is a deficit of refinery gas, Liquefied Natural Gas (LNG) can
be imported to use as a make-up. The cost for producing steam is dependent on the
amount of purchased LNG to fuel the steam boilers. Work for pumps and compressors
can be obtained either by electricity (motor mode) or by steam (turbine mode). This
possibility to switch energy source affects the steam balance of the refinery. Furthermore,
electricity and LNG prices affect the choice of driver mix that minimizes the refinery
utility costs. Thus there is a clear need for a model of the steam utility system that can
be linked to a tool for optimizing the operating cost for pumps and compressors.
The basis of this project was a model developed in Aspen Utility Planner in a previous
master thesis. This model has been further developed and improved in this project to
become easier to use and run from an Excel interface. Furthermore, the model has been
improved to better represent the steam network at Preemraff Lysekil.
After an investigation of key variables, such as the production and consumption of steam
in the different production units at the refinery, the model was validated against mea-
surements from different operational scenarios.
Steam system simulations can be run through the Excel interface. The Aspen Utilities
Planner simulation environment is only required for development of the steam system
flowsheet configuration. The model was tested for a number of representative operating
situations, and it was concluded that the model provides reliable results for stable op-
erating conditions and also provides results that are within the acceptable error margin
for unstable operational situations i.e. when parts of the refinery are shut down, but
for these cases the reliability of the model decreases. The optimization function is work-
ing and provides solutions that reduce the estimated utility cost. Further investigations
should concern investigation of steam system balances during operating situations when
parts of the refinery are shut down.
Keywords: Steam system, Optimization, Utility cost, LNG, Refinery gas, Aspen Utilities
Planner.
v |
Chalmers University of Technology | 1
Introduction
1.1 Background
The Paris Agreement signed in 2015 aims to respond to the threat of climate change
by limiting the risk of global temperature to well below 2 ◦C above pre-industrial levels.
This will require many different types of action, including reduced energy consumption.
In Sweden, industry had a consumption of 140 TWh out of a total final usage of 370
TWh [1] in 2015. According to Preem’s environmental report from 2015 [2], Preem had
a total fuel usage of 6.4 TWh, this is the contribution from the use of fuel gas, LNG and
coke from the cracker unit. Preem AB represents 80% [3] of the Swedish refining capacity,
corresponding to 345 000 barrels per calender day [3]. Specific numbers from the agency
"Naturvårdsverket" show that Preem Lysekil had CO emissions of roughly 1.4 million
2
tonnes in total in 2016 [4].
Steamisofthehighestimportanceinordertokeeptherefineryrunning. Mostofthesteam
is produced in boilers that mainly use residual gases (light non-condensed components)
from the process, so called refinery gas, as a fuel. However, if needed or if there are
economic advantages, make up gas to the fuel gas system can be obtained by purchased
liquefied natural gas (LNG). There is sometimes an excess of steam, which is because the
refinery gas has no market value and too much flaring of refinery gas is not allowed due to
environmental regulation. Therefore, in periods when there is an excess of refinery gas, it
is better for the refinery to produce an excess of steam. Some venting will always occur
since producing exactly what is required is operationally difficult. In the refinery steam
system, very high pressure (VHP) steam is generated and let down to lower pressure levels
and, if necessary, the amount that is not needed is vented to the atmosphere causing an
energy loss. Other than letting down steam through the let down valves between each
header level, there are machines like pumps and compressors that can be driven by steam
turbines or electric motors, depending on steam availability and the economic trade-off
between the electricity cost for motor drive and the fuel cost for steam production. The
possibilitytoswitchthemachinestobedrivenbysteamturbineinsteadofmotordrivecan
be utilized to increase the steam demand and reduce steam loss from venting. Selecting
which pumps and compressors that should be driven by motors and steam turbines is a
complex combinatorial problem, which means that an optimization model is useful for
making the operation of steam utility system as economically efficient as possible. The
amount of steam that needs to be produced will also affect the fuel gas system balance.
A more detailed description of the steam network with its main components is described
in Section 2.2.
A steam model that reflects the real system well and therefore can be utilized by the
1 |
Chalmers University of Technology | 1. Introduction
refinery to improve their economic and environmental performance. For this reason,
Subiaco [5] developed a steam model representing Preem’s steam system. However, the
modeldevelopedbySubiaco[5]isinneedofadditionalvalidationandfurtherdevelopment
in order to be useful as a decision support tool for Preem staff.
1.2 Aim and objectives
The aim of this work is to improve and develop the steam system model created by
Subiaco [5] into a decision support tool for operational optimization at Preemraff Lysekil.
To achieve the aim, the objectives are to:
• validate key variables and improve the model so it better reflects actual operation
of the refinery. This will give results that are more reliable and closer to values
measured at the refinery.
• make the optimization mode functional, verify its reliability and make it easy to use
through the Excel interface,
• develop a user interface for convenient updates of model data according to the most
recent process data values,
• develop a basic mapping of the marginal changes in the LNG fuel system due to
changes in operation of the steam system,
• developabetterfuelgassystemmodelwhichwillmaketheestimationoftherefinery
gas and the LNG consumption more reliable.
The work has been carried out in sequences according to the milestone seen in Figure 1.1.
Figure 1.1: Main step in project execution.
2 |
Chalmers University of Technology | 1. Introduction
1.5 Literature review on operational optimization of
utility systems
The focus of this literature review is to obtain an understanding of the concepts behind
the program that is used, present information about actual implementation of similar
models at process plants and what are the practical aspects of a steam network that can
be included in a theoretical model.
Partly based on the work by Papoulis and Grossmann [6] Micheletto et al. [7] in Sao
Paulo in Brazil at RECAP refinery formulated a MILP problem of the refinery’s utility
system that involved energy and mass balances, the operational status of refinery units
connected to the steam system and steam consumption/production of these units. The
modelwasabletodecreasetheoperationalcostsfortherefinerybyupto10%byproviding
an optimal configuration of the setting of units and also by identifying steam losses and
inefficient units. The model was integrated into RECAP’s database in order to plan on
using the utility system efficiently. This shows that a MILP solver has been successfully
integrated into a real operating plant with decreased operational costs and also that it
can interact with operational data from a refinery.
As one of three improvements to their plant model, Zhang and Hua [8] suggested to
incorporate a MILP model of the utility system in the complete plant system in order
to improve energy efficiency. The other two improvements focusing on consumption of
the units in the process and balancing the steam, fuel gas and fuel oil for the plant. The
model approach including above improvements were implemented in a real industry. This
shows that a MILP model of a utility can be integrated with the remaining parts of a
process. It also shows that the model in this paper is well built as it also is subdivided,
thus decreasing complexity of the flows.
The comparison by Bruno et al. [9] between MILP and MINLP showed that using fixed
variables for variables such as pressure and efficiency (MILP) gave results which were
consideredinfeasiblecomparedtotheresultsobtainedfromtheMINLPsolver. Thisisdue
tothefixedoperatingparametersinaMILPmodel,i.e. theflexibilityoflessfixedvariables
give a more optimal outcome. In the MINLP, model variables such as temperature and
pressure are not fixed and therefore the system becomes nonlinear. According to Bruno,
the ability of MINLP models to handle such non linearities gives more reliable results
than a MILP model.
A practical approach for making steam systems more efficient and track steam consump-
tion was suggested by Aegerter [10] and Bickham and Wadel [11], where the focus is more
on practical issues such as keeping boiler efficiency high, turbine operation and main-
tenance of equipment in the process. Aegerter [10] argues that, for example, a faulty
valve can leak through around 4500 kg/h of steam. The importance of investigating the
performance of equipment and pipes and how this is handled by operating staff is high-
lighted. In this report, the practical aspects for making a steam system more efficient and
investigating on the leakage will be taken into the theoretical MILP model.
4 |
Chalmers University of Technology | 2
Steam network at Preem Lysekil
The complex steam network at Preem refinery in Lysekil consists of several components
and variables that affect the operation of the network. A description of the main compo-
nents is provided together with examples of research connected to Preem refinery.
2.1 Research connected to the studied oil refinery
In the previous research by Riccardo Subiaco involving the Preem refinery and the steam
network a model for simulation and optimization of the steam network was developed
[5]. This model is also the foundation and starting point for this thesis. Cristina Murcia
Mayo [12] studied in her MSc thesis computer based analysis tools for handling of data
for industrial energy systems analysis. Also CIT Industriell Energi AB has, as part of a
research collaboration between Preem and Chalmers, conducted pinch analysis study of
the refinery in Lysekil [13].
Studies being performed currently include a project by Ph.D candidate Sofie Marton
who is using the refinery in Lysekil as a case study for heat integration. The study will
investigate a number of retrofits for heat exchanger within the refinery. The retrofits are
being investigated in the perspective of operability connected to the changes in the heat
exchanger network [5].
Examples of other research projects that have been conducted in collaboration between
Chalmers and Preem regarding heat integration [14], bio-refinery with biomass feedstock,
chemical looping combustion, automation of heat integration project and catalytic reac-
tions regarding bio-oils. For short descriptions of these projects, see [15].
2.2 Description of the steam system
As mentioned in Section 1.1, steam is one of the most important hot utilities in an oil
refinery [16]. It is used both as a heat carrier and a source for mechanical work in the
refinery. Steam is in this particular plant produced in steam boilers, heat recovery steam
generators (HRSG) and process coolers. The steam network at the refinery consists of
four pressure levels, also called headers; very high pressure (VHP), high pressure (HP),
medium pressure (MP) and low pressure (LP) [17]. Equipment that works between the
pressure levels such as pumps, compressors and blowers are units that can be set in two
different modes, motor or turbine. For motor mode, electricity is the source of power
and for turbine mode, steam is the source of power. These two modes are not used
5 |
Chalmers University of Technology | 2. Steam network at Preem Lysekil
simultaneously, either the unit is motor driven or steam driven. Figure 2.1 illustrates the
general overview of the steam network in the refinery.
Figure 2.1: The overview of the refinery steam network [17].
The HRSG:s and boilers are the main producers of VHP, there are three boilers and
two HRSG:s [5]. HRSG:s produce VHP steam by recovering heat from flue gases. For
the operation of boilers, different scenarios for VHP steam can be identified depending
on which time of the year it is. The boilers use mainly refinery gas to produce steam
and the refinery gas mainly consist of non-condensed lighter substances. The amount
of refinery gas obtained will partially depend on the ambient air temperature. In the
summer when the air has high temperature, a smaller amount of lighter substances can
be condensed compared to the winter. This results in a greater amount of refinery gas.
Thus, in summer, VHP steam will be produced mainly by using refinery gas for steam
boilers and the HRSG:s are usually not in operation. In the winter, on the other hand,
when the ambient air temperature is low, the cooling system operates more efficiently.
The amount of refinery gas obtained will generally be less compared to summer. In this
situation, the energy demand for the production of VHP steam cannot be covered with
refinery gas only. Consequently, purchased LNG is used as a make-up fuel. This leads to
6 |
Chalmers University of Technology | 2. Steam network at Preem Lysekil
the consideration of how to balance the energy demand. A decision between importing
LNG and keep producing the same amount of steam or going over to motor driven units
is crucial. In the future, the trade-off between LNG and electricity may very well be
important all year around, depending on prices and emission restrictions.
The HP steam originates from process cooling and a mixture of steam that is throttled
from excess VHP steam. Steam at MP and LP level both have inflows of steam in
similar ways, de-pressurized steam from higher pressure levels that has performed work
in turbines, from an excess of steam and is throttled and also from process coolers.
As the condensate leaves the LP level, parts of it can be recovered and reused in order
to reduce the demand for make-up water. The main lines of the steam system are the
headers which are branched out over the entire process (except HP, which is at the new
hydrogen producing unit) [5] and then distributes the steam to the consumers i.e. mainly
steam heaters. As there are four different pressure levels there are four main headers, one
foreachlevelextendedalongtheentireplantandtheyareconnectedthroughthethrottles
and turbines that act as pressure sinks. Unlike other pressure levels, the HP steam header
is a local header which is used only to supply steam at the newest hydrocracker unit [17].
The trade-off between electricity price and the price for LNG is of importance during the
colder period of the year since the amount of refinery gas is usually not enough to cover
the demands of the boilers during this part of the year. During the cold period of the
year the HRSG´s produces larger amounts of steam and the energy content of the hot
flue gases produced within the refinery is recovered for steam production thereby making
use of energy that otherwise would remain unused.
The environmental regulations also come in as a variable for how to balance the steam
production. The combustion of fuel gas through flare stack is one way to reduce the excess
steam, however, over-flaring can violate environmental permits. Hence, flaring should be
kept within its regulation and consequently the use of fuel gas should be maximized.
2.3 Main components of the steam system
There are several units in the refinery that interact with the steam system. So, under-
standing of how they interact and affect the steam system is important to follow the
procedure of this report.
2.3.1 Steam header
The steam headers are spread out around the refinery, connected to both producers and
consumers. Pressure and temperature at the headers are considered constant, however,
since it can be some distance between a producer and a consumer these statements are
not completely true but the differences are small enough for the assumption to hold. As
described in Section 2.2 there are four main headers, of which one is purely for one specific
area (810) of the refinery.
7 |
Chalmers University of Technology | 2. Steam network at Preem Lysekil
The number of steam consumers increases as steam goes to the lower pressure headers.
At the VHP header the steam is mainly used to obtain mechanical work for high power
demanding units and the use of VHP steam as injection steam is small in relation to the
total steam flow at this header. The HP level header is confined to a specific area which
is relatively new, and the HP level measurement system is more reliable and tracking the
steam easier as this subsystem is less complex. The MP level is similar to the LP level
and it is also connected to a high number of heat exchangers. Thus the tracking of steam
is more difficult at the MP and LP levels due to the greater number of connections to heat
exchangers, direct steam users, and other unidentified and unknown steam consumers.
Steam that is extracted from turbines and let-down valves is super-heated. This means
that the temperature is above the saturation temperature. In order to keep the steam
at the steam headers at saturation temperature, de-superheaters are used. These units
inject water at lower temperature that cools down the steam to the desired temperature.
2.3.2 Steam tracing and steam traps
Steam tracing is the heating of pipes in the process and heating of tanks. The reason for
this is to maintain desired temperature of the fluid inside the pipe or tank so that appro-
priate flow properties are maintained and the fluid can be easily pumped and transported
without too high friction losses. The heating demand for steam tracing depends partly
on the season of the year and it is difficult to track since there are no flow measurements
and the documentation is lacking.
Steam traps are positioned along the steam headers meant for removing condensed steam
that would otherwise accumulate within the pipes and affect the steam quality and also
be a cause of corrosion. The removal is different from trap to trap, in some traps the
condensate is let out to the ground while in others the condensate is expanded to the
header below.
2.3.3 Let-down valves
The let-down valves are directly connected to different headers and are used to allow
make-up steam from one header to the header below and also to avoid the ventilation of
steam at high pressures in case of overproduction. Flow rate equations for these valves
can be obtained from Preem´s operational system as a function of the valve opening.
These equations are of importance since the flow measurement is not always reliable.
By comparing the flow measurement with the valve opening the reliability of the flow
measurement can be improved.
2.3.4 Switchable drives
At the refinery there are pumps, compressors and blowers that transport different media.
These units are driven using either electricity by motor or steam expansion through tur-
bines. This is designed so that a unit, for example P-3204 has two units A and B where
one is motor-driven and another one is a turbine-driven. The setup design can be seen in
Figure 2.2.
8 |
Chalmers University of Technology | 2. Steam network at Preem Lysekil
Figure 2.2: The design setup for pumps/compressors, adapted from [5].
However, there are some units that have the setup of two turbine-driven and one motor-
driven pump, then a second turbine unit can be added to Figure 2.2. This second turbine
can be considered as a back-up unit, i.e. this setting is implemented for units that are
fundamental for the operation of the refinery. Examples are the pumps that feed steam
producers with water.
Furthermore, there are some units that only utilize either the turbine or the motor during
extreme cases. For instance, the blowers for the steam boiler only utilizes the motor unit
during start-up since it takes its steam from the boiler which it is connected to. The
opposite example is that some units only use the motor since the response is faster. In
such cases, the pump is only used partly for keeping a level, in situations like this the
motor is the one that is operational and the turbine is started manually when there are
longer operational deviations.
The refinery does not measure all the steam that goes through every turbine. However,
the current that is used for the motors is measured. As further explained in Section 6.1.3,
this, can be used to determine the power demand of pumps and compressors.
2.3.5 Steam boilers and HRSG:s
There are three steam boilers at the refinery. Due to operational security at least two
boilers always need to be in operation. If only one boiler would be in operation and there
would be an emergency shutdown of that boiler, this would lead to a shutdown of the
whole refinery due to a failure from having insufficient steam. The boilers are the most
flexible steam producers, HRSG:s are limited by the amount of hot flue gases and the heat
exchangers are limited in the same way by the heat content of the hot process stream.
Fuel to the boilers can be imported to fit the need for steam production.
9 |
Chalmers University of Technology | 2. Steam network at Preem Lysekil
Theboilersoperatewithinanintervalofsteamproduction. Operationattheupperlimitis
unusual, where an operation close to the lower limit is quite common. There are examples
of situations when the boilers, in reality, are operated below the nominal minimum load
limit. Thus the value of the lower limit will be of special interest for the model validation,
as further discussed in Section 6.1.2.
The HRSG:s have, like the boilers, upper and lower limits for steam production, however
if they are shut down they still produce approximately 1-2 t/h of steam as the water is
circulated to avoid over-heating and the flue gas is mainly by-passed, the only way to
completely stop it is to block the incoming water.
2.3.6 Fuel gas system
Refinery gas mainly consists of light components that are difficult to condense, there are
around 20 producers of refinery gas and the majority are vessels and towers. The number
of consumers are around 25, mainly furnaces. The production of internal refinery gas
partly depends on the ambient temperature, but according to Preem staff there are other
factors that affect it as well and it is too simplified to relate the production to outdoor
temperature only.
The production of refinery gas is measured and there are also measurements after the
refinery gas has been mixed with the imported LNG. After mixing, the measurements are
extensive, the variables that are measured are, for example, density and heating value
etc. which are controlled.
Preem can, as mentioned in Section 2.2, import LNG when needed, the mix of LNG and
refinery gas is the fuel gas which is incinerated in the boilers to create heat for steam
production. At times when there is no need for LNG import, fuel gas is only pure refinery
gas. At times when the fuel gas is pure refinery gas the amount of produced steam
from the boilers cannot be decreased since the refinery gas cannot be stored nor flared
excessively due to environmental permits as described in Section 1.1.
10 |
Chalmers University of Technology | 3
Aspen Utilities Planner
Aspen Utilities Planner is a part of the Aspen Energy & Utilities Optimization tool,
which is an equation oriented tool designed to simulate and optimize utility systems. It
also handles economical calculations that are connected to the utility system. Different
kinds of utility systems such as power, fuel and steam can be handled in the Aspen
Utilities [18]. Like other Aspen programs such as Aspen HYSYS or Aspen Plus, Aspen
Utilities Planner creates a flowsheet in which the process is modelled by using blocks that
represent different units in the process and also simulates these units’ behaviour. In the
program there are two modes that can be utilized during steady state simulation, these
are: scenario and optimization modes.
In Aspen Utilities Planner there are three kind of variables that can be used; fixed vari-
ables, free variables and initial variables.
• Fixed variables are the input parameters needed to be specified. They can be kept
constant, for example, the specifications of equipment and efficiencies but some
of them i.e. temperature and pressure can be changed by users to test different
operating conditions.
• Free variables are the unknown variables which will be adjusted when simulating
the model, if the problem is feasible.
• Initial variable are only used for dynamic simulation which is not of interest here.
3.1 Scenario mode
In this mode, the number of fixed variables are less than in optimization mode since the
focus is to mimic a certain operational situation and setting in a time period or time
point. This mode is a good step before using the optimization tool, this since the results
can be used to troubleshoot a model under development. By comparing output values
to measured values from Preemraff Lysekil, a hint of how accurate the model is can be
obtained, assuming the measured values are reliable. Once the model produces reliable
results that are consistent with measured ones the model can be used in optimization
mode.
The setting of the units (pumps, compressor etc.) is fixed, as are temperature and pres-
sure at the steam headers, power demand of the units (pumps, compressor etc.), some
efficiencies for example for boilers and motor driven pumps, as well as the consumption
and production of steam including the boilers. Flow through valves between the headers,
LNG flow and water make up to the system are the variables that are used by the solver
11 |
Chalmers University of Technology | 3. Aspen Utilities Planner
to solve the mass balances in this mode. Steam flow through units are free variables
although constrained by temperature and pressure at the headers, thus making previous
mentioned flows the variables that will be used by the solver.
3.2 Optimization mode
The solver used by the Aspen Utilities Planner is a Mixed Integer Linear Programming
(MILP) solver. MILP is a mathematical optimization program used for problems with
linear constraints, objective functions and a mix of continuous and integer variables.
In this mode there are more free variables than in scenario mode, this is for the solver to
have more degrees of freedom. Examples of fixed variables during optimization, some of
which are the same as in scenario mode, are; temperature and pressure at the steam head-
ers, power demand of the units (pumps, compressor etc.), some efficiencies for example for
boilers and motor driven pumps and the consumption and production of steam excluding
the boilers. Example of free variables are flow through valves, steam flow through units
although constrained by temperature and pressure at the headers as described in Section
3.1 and boiler steam production.
The solution to the optimization model should be a more cost effective operational set-
ting of the switchable drives that optimizes the trade-off between use of electricity and
production of steam to minimize total utility costs. It is not as simple as to say that a low
electricity price means that all units should be motor driven or vice versa, the optimal
solution will probably include a mix of turbine and motor driven units. For the program
to be able to solve such a complex problem, the number of fixed variables needs to be
smaller than for scenario mode. Instead, creation of inequality and equality constraints
is also needed. An example of constraints in the model are the operational settings of the
units in the model, which can be either:
• Available,
• Must Be On, or
• Not Available.
Here, "Available"referstoaninequalityconstraintwhile"MustBeOn"and"NotAvailable"
are equality constraints. Further examples of constraints are minimum and maximum val-
ues for units. These constraints will keep the solver within realistic values, preventing it
from yielding values that are nonphysical. However, the constraints regarding steam
producing units such as internal coolers and the operational mode for pumps and com-
pressors depend on the operating scenario, hence, the constraints need to be edited by
using so-called data editors, see Section 3.2.1.
12 |
Chalmers University of Technology | 3. Aspen Utilities Planner
3.2.1 Data editors
In the data editor, there are basically three kinds of data that the user needs to define in
AspenUtilitiesPlanneri.e. demand, availabilityandtariffs. Demandisaconstraintgroup
of utility supplied or demanded for each equipment. Availability is used to set the on/off
constraints as well as the minimum and maximum possible value for each equipment.
Demand and Availability profiles can be found in the ’Profile’ database. Tariffs are the
purchasing or selling price for each utility type that are used in the plant and can be
found under ’Tariff’. Besides these three editors, there is a more advanced editor called
Demand Forecasting but this capability was not used in this project. Figure 3.1 shows
the default editor interface in Aspen Utilities Planner.
Figure 3.1: Default Aspen Utilities Planner Data Editors.
AspenUtilitiesPlannerhasadefaultformatforeachdatatypeandisstoredasaMicrosoft
Access database, Tariff data is stored as a TariffData.mdb, Demand and availability data
are stored as a Profile.mdb. There is another database used by Aspen Utilities Planner
named Interface.mdb. The interface is used when the optimization mode is run and will
collectdatafromotherdatabasesandstoretheresults. Figure3.2displaystherelationship
between the various editors and databases.
13 |
Chalmers University of Technology | 3. Aspen Utilities Planner
Figure 3.2: Overall relationship between various data editors and corresponding
databases, adapted from [18].
Apracticalproblemrelatedtotheoptimizersolvercanbethatthesimulationisunfeasible.
Here, the error diagnostic function, which are presolve error checking and error tracking,
can be used to detect the cause of the problem.
The presolve error checking function is used to identify if there are infeasibility errors
present, for example, when a variable has its maximum bound smaller than the minimum
bound. This kind of feasibility problem is easy to tackle.
On the other hand, there could be other types of error that are much more difficult to
detect. This kind of unfeasible error takes place when the mass balance for a process block
does not agree and cannot be detected by the presolve function. Another error diagnostic
function called "error tracking" is used to deal with the problem. By introducing an
additional variable to each balance equation which is minimized in the objective function,
the problematic equations and the corresponding blocks will be shown in the message
window.
3.3 Microsoft Excel interface
With the use of Aspen Utilities Planner add-ins ’Utilities340’, it is possible to connect
the flowsheet from Aspen Utilities Planner to Microsoft Excel spread sheet. The users
can develop their unique spreadsheet to send inputs and retrieve results, shown in Excel
interface. An explanation on how to connect and use Aspen Utilities Planner - Microsoft
Excel interface from scratch can be found in ’Aspen Utilities User Guide V8.8’ [18].
14 |
Chalmers University of Technology | 4. Original model of the refinery steam system
In Figure 4.1, all producers and consumers of steam to, from and in between the steam
headers have been modelled. For convenience, consumers and producers of steam that
are connected to the same header have been lumped depending on category, for example
all the internal process heating at each header is represented by a single condenser, all
internal process cooling is represented by an evaporator. Steam consumption that leaves
the system, i.e. injections and similar, are represented by a demand block. Also the units
that works between the different headers are lumped in order to make overview easier,
one pump in Figure 4.1 can represent a number of pumps in reality.
Another practical feature of the model is that the units are lumped together not only
depending on between which headers the unit is working, but also according to Preem’s
own unit classification. This makes it easier for example to trace the steam flow in a
specific point in the physical process.
Parts of the system representing steam consumption, such as leakage and other steam
consumers for which there is no measurement, have also been aggregated. These are
represented in Figure 4.1 as a heat exchanger and a steam demand in the green square
at the lower right-hand corner. These two flows are connected to the water balance and
affects the variable representing the make-up water to the system.
The fuel gas system in the original model created by Subiaco [5], consists of refinery gas
supplier, LNG supplier and fuel header which contains the mixed gases of refinery gas and
LNG so-called fuel gas. The fuel gas system was originally modelled assuming a constant
LHV of fuel gas and LNG. At Preemraff Lysekil, the composition of the mixed fuel gas is
measured by on-line gas chromatograph and after that the LHv and density is calculated
based on the composition. The LHV of the fuel gas needs to be converted into mass
basis by using its density before feeding to the model. LHV of LNG is not measured but
instead its composition is measured, therefore the LHV of LNG can also be calculated.
In the original model created by Subiaco [5], the composition of fuel gas at the header
feeding fuel to the boilers was calculated assuming a fixed %LNG by molar composition
for a given scenario. The LHVs of both refinery gas and LNG were specified as fixed
values as can be seen in Table 4.1.
Table 4.1: LHVs of refinery gas and LNG from original model.
Fuel LHV[MJ/kg]
Refinery gas 37
LNG 45
With the LHVs, molecular weights of both refinery gas and LNG, and the percentage of
LNG in the fuel gas together, the heat provided from the relationship between the refinery
gas and LNG in the original model created by Subiaco [5] was calculated from Equation
4.1.
16 |
Chalmers University of Technology | 4. Original model of the refinery steam system
n = %LNG×(n +n )
LNG LNG re
!
Q Q Q
LNG LNG re
= %LNG× +
LHV ×MWT LHV ×MWT LHV ×MWT
LNG LNG LNG LNG re re
(4.1)
Regarding variables in the original model, they are built up similarly to the descriptions
in Sections 3.2 and 3.1. The number of fixed variables depends on the type of simulation.
Variables that are always fixed are temperature, pressure, steam production from process
cooling and steam consumption for injection and process heating. During scenario mode
simulation, the operational setting of pumps and compressors are also fixed, as well as
the steam production from the boilers. In optimization mode, the operational setting of
pumps and compressors can be set freely and also the steam production from the boilers,
within the constraint boundaries. For a more detailed description of the original steam
system model, see [5].
For the original model created by Subiaco [5], an Excel interface was available for running
the model and structuring the resulting output. However, running simulations from the
Excelinterfacewaslimitedtoscenariomodesimulationrunsonly. Theoptimizationmode
could not be operated from the Excel interface. In this project, Excel spreadsheets where
new scenarios could be added and a simplified flowsheet of the steam network existed and
have been further developed.
17 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.