University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
ADE | Chapter 7 – Fluid Transients and ANN for Burst Detection and Identification
hand, the maximum median error in the prediction of the burst size was found at the
beginning of the pipeline, when the burst was located at only 0.36 m from the upstream
end of the pipeline and most of the median size predictions were between −0.5 and
+0.5 mm. Once the model development stage was completed, meaning that the two
ANNs had been trained and tested, new numerical traces were generated to be tested
at the application stage of the proposed technique.
7.5.2 Application Stage
To assess the performance of the burst detection and location ANNs and to
demonstrate the application stage of the proposed technique, numerical traces
corresponding to different bursts were generated to replicate Step B.1. This section
presents the results for one of these traces as an example. The obtained transient head
trace is presented in Figure 7-8. This transient head trace corresponds to a burst located
at 125.75 m along the pipeline with an orifice burst size of 56 mm. Figure 7-8 also
shows the length of a typical time window that would be selected at Step B.2. For the
dimensions of the pipeline described in Figure 7-2, the length of the time windows is
2.25 s.
Figure 7-8. Numerical transient head trace for burst located at 125.75 m along
pipeline and burst size of 56 mm.
Once the first time window has been selected, the burst detection analysis is conducted
by downsampling the time window to match the sampling frequency of 256 Hz (Step
141 |
ADE | Chapter 7 – Fluid Transients and ANN for Burst Detection and Identification
B.3) and by processing this time window through the burst detection ANN (obtained
at Step A.7). The results from this analysis are presented in Figure 7-9 for the first
1,210 time windows of the full transient head trace of Figure 7-8.
Figure 7-9 presents the category predicted by the burst detection ANN for each time
window from Figure 7-8. In addition, this plot includes the correct category in black
for comparison. This figure demonstrates that the burst detection ANN classifies
correctly all the time windows that correspond to Category N, and for the windows
that correspond to Categories Ab-I and Ab-C, there is a short lag (of four time
windows maximum) before the burst detection ANN correctly recognizes a time
window. This classification lag is not significant and would represent a lag of 0.016 s
in the detection of a burst.
Figure 7-9. Prediction of head condition categories in burst detection ANN for a
numerical transient head trace.
Once a time window has been identified as Category Ab-C for the first time, the
consistency test (over 𝐿/𝑎 seconds) is triggered (Step B.4 in Figure 7-3). As an
illustration, Figure 7-9 includes a rectangle that encloses the 230 time windows that
belong to the consistency test and have been processed in the burst identification ANN
at Step B.6. The results for the prediction of the location and size of the burst are
presented in Figure 7-10.
142 |
ADE | Chapter 7 – Fluid Transients and ANN for Burst Detection and Identification
Figure 7-10. Distribution of prediction in burst identification ANN for burst (a)
location; and (b) size for numerical transient head trace.
Considering that 230 time windows are included in the consistency test (covering 𝐿/𝑎
seconds of the transient head trace), the results from the burst identification ANN
include 230 location and size predictions for the burst registered in the transient head
trace. Figure 7-10(a) shows that the obtained distribution of the burst location
predictions is between 111.4 and 135.5 m when the real location is 125.75 m. This
plot also shows that the obtained distribution includes predictions before and after the
real burst location and that the median prediction is only 0.85 m away from the real
burst location (which represents 0.085% of the total length of the 1,000-m-long
pipeline). Figure 7-10 (b) presents the distribution of the predictions for the burst size,
from which it is possible to determine that the median size of the burst is predicted to
be 0.17 mm (an error of 0.30% from the actual burst size of 56 mm).
To complete the application stage, a burst verification analysis was conducted. Using
the median predictions for the burst location and size, a transient head trace is
generated and compared to the transient head trace presented in Figure 7-8. This
comparison is shown in Figure 7-11.
143 |
ADE | Chapter 7 – Fluid Transients and ANN for Burst Detection and Identification
Figure 7-11. Burst verification analysis for numerical transient head trace.
From this figure it is possible to observe that the two transient head traces are quite
similar, even considering that the initial head variation is different. In addition, the
NRMSE was calculated, obtaining a difference of 1.1% between both traces. This
proves that the proposed methodology results in accurate predictions of the burst
location and size. Figure 7-6 and Figure 7-7 also demonstrate that the methodology
yields good results when applied to transient head traces not used for the training of
the ANNs.
7.6 Experimental Results
A series of experiments in the Robin Hydraulics Laboratory of the University of
Adelaide were conducted to validate the proposed methodology for detecting and
identifying bursts in pipelines using ANNs. The layout of the experimental pipeline
system replicates the system configuration presented in Figure 7-2. The upstream end
of the pipeline is connected to a pressurized tank, and the downstream end has a closed
inline valve. The characteristics of the pipeline are shown in Table 7-2.
To validate the methodology described in Figure 7-3, all the steps of the model
development stage were replicated for the pipeline in the laboratory. The architecture
of the ANNs and the number of samples to create the input dataset for the ANNs were
144 |
ADE | Chapter 7 – Fluid Transients and ANN for Burst Detection and Identification
not modified; however, new transient head traces were numerically generated taking
into consideration the pipeline characteristics presented in Table 7-2.
The fast opening of a side-discharge solenoid valve was used to model the occurrence
of a burst in the pipeline. Therefore, the numerical modeling of the transient head
traces included a typical opening curve for a solenoid valve with an opening time of
10 ms. The downsampled frequency selected was 5 kHz considering that the pressure
head in the experiment was measured at the downstream end of the pipeline using a
PDCR 810 pressure transducer (manufactured by Druck Limited Leicester England)
with a 10-kHz sampling rate.
Once the numerical transient head traces were generated, the time windows were
obtained, classified, and used to train and test the burst detection and burst
identification ANNs. Because the same procedure was explained for the numerical
results, the results of the training and testing of these ANNs are not included here.
However, considering that stochastic gradient descent algorithms were used for the
ANN training, each time that this process is conducted, the resulting weights are
different. Therefore, both ANNs were trained multiple times (without modifying its
architecture) to assess the robustness of the obtained results. In the case of the burst
detection ANN, the results were not sensitive to changes in the ANN weights, so only
the results from one training attempt are presented. For the burst identification ANN,
results will be presented in what follows in this section for three of these training
processes. Each training process for the burst detection ANN took approximately 3 h,
while the training of the burst identification ANN took 15 h on average.
The transient head trace obtained from the experimental test is shown in Figure 7-12.
This trace corresponds to a burst (modeled as the opening of a side-discharge solenoid
valve) located 11.06 m downstream of the pressurized tank and has a size of
approximately 2.0 mm. The initial pressure head before the occurrence of the burst
was 29.48 m. Figure 7-12 also presents the length of the time windows for this
pipeline. The length selected was 2.5𝐿/𝑎 seconds, which corresponds to 0.072 s.
145 |
ADE | Chapter 7 – Fluid Transients and ANN for Burst Detection and Identification
Figure 7-12. Experimental transient head trace.
The application stage was validated on the transient head presented in Figure 7-12,
following an individual analysis of each time window in sequence. Results from the
burst detection analysis (Step B.4 in Figure 7-3) are presented in Figure 7-13. This
figure presents the category predicted by the burst detection ANN for the first 577
time windows of the complete transient head trace. In addition, the correct category
for each of these time windows is included with the objective of illustrating where the
burst detection ANN misclassified a time window.
As can be seen in the figure, most of the time windows are correctly classified, except
for one time window that belongs to Category N and was classified as Category Ab-I.
In addition, some other time windows were classified in Category N, but belong to
Category Ab-I (showing a time lag in the ability of the burst detection ANN to identify
the occurrence of a burst), and two time windows that belong to Category Ab-I were
classified into Category Ab-C (not visible in the plot).
146 |
ADE | Chapter 7 – Fluid Transients and ANN for Burst Detection and Identification
Figure 7-13. Prediction of head condition categories in burst detection ANN for an
experimental transient head trace.
This figure also shows the time windows that were considered in the consistency test.
A total of 144 time windows were classified as Category Ab-C and so are enclosed in
the rectangle in the figure. These 144 time windows were processed by the three
different burst identification ANNs, and the distribution of the predictions for the burst
location and size are presented in Figure 7-14.
Figure 7-14. Distribution of predictions in burst identification ANN for (a) burst
location (along 37.24-m pipeline); and (b) burst for experimental transient head
trace.
Figure 7-14(a) shows that the three burst identification ANNs provide different
distributions of results, but for the three cases, the median predicted location is within
1.50% of error in terms of the total length of the pipeline (37.24 m), as shown in Table
7-3. The closest prediction to the real location of the burst was obtained for the third
burst identification ANN in which the burst location was predicted to be only 0.06 m
away from the real location. In a similar way, Figure 7-14(b) presents the distribution
in the predictions of the burst size. The three burst identification ANNs predicted
147 |
ADE | Chapter 7 – Fluid Transients and ANN for Burst Detection and Identification
different burst sizes, resulting in a median error between 6% and 16% of the true burst
size. Although these results are less accurate than the burst location results, it is
important to note that the largest median error corresponds to an absolute error of only
0.32 mm. In addition, the computational time required to obtain these results was only
a few seconds.
Table 7-3. Predicted Results by Burst Identification ANNs. L =37.4 m.
T
Median Median
Burst True True Median Median
Location Size NRMSE
Identification Location Size Location Size
Error Error (%)
ANN (m) (mm) (m) (mm)
(%) (%)
1 10.63 2.19 1.15 9.50 2.58
2 11.06 2.00 10.55 2.32 1.37 16.00 5.61
3 11.00 2.12 0.16 6.00 2.83
Lastly, a burst verification analysis was conducted. Three different transient head
traces were generated taking into consideration the prediction of the three burst
identification ANNs and they were compared to the experimental head trace presented
in Figure 7-12. The comparison is shown in Figure 7-15(a), and the obtained values
for the NRMSE are presented in Table 7-3. There is significant agreement between
the transient head trace obtained in the laboratory and the numerical transient head
traces generated from the predictions of the burst identification ANNs in terms of the
occurrence of the head drop and recovery due to the arrival of the burst wave to the
measurement point. This is due to the accurate prediction of the burst location.
Figure 7-15. Burst verification analysis: (a) numerical transient head comparison;
and (b) results after burst prediction adjustment.
148 |
ADE | Chapter 7 – Fluid Transients and ANN for Burst Detection and Identification
However, differences are visible in the magnitude of the head drop for the three
predictions. These differences are due to the error in the prediction of the burst size
given that the magnitude of this head drop is directly related to the burst size.
Following the proposed methodology, a burst prediction adjustment was applied
taking into consideration the discrepancy between the transient head traces in Figure
7-15(a) and the fact that the values for NRMSE are larger than the selected threshold.
Taking into consideration the prediction location obtained from the burst identification
ANNs, a series of transient head traces with different burst sizes were generated
covering the complete range of possible burst sizes (ranging from 0.4 to 3.5 mm in
increments of 0.1 mm). For all of these transient head traces, the NRMSE was
calculated when compared to the experimental transient head trace. Figure 7-15(b)
shows the transient head trace with the lowest NRMSE, which corresponds to a burst
size of 2.0 mm (obtaining an error of 1.47%). It is important to highlight that only 31
transient head traces were generated for this adjustment, which does not represent an
impractical computational time. Thus, the proposed technique was able to locate and
characterize the occurrence of a burst in the laboratory with enough accuracy based
only on the interpretation of the measured transient head trace.
7.7 Conclusions
This paper presents a novel technique to detect and characterize the occurrence of
bursts in pipelines by merging the use of fluid transient waves and artificial neural
networks. A complete methodology was described and divided into two main stages:
model development and model application. The model development stage includes the
generation of a transient head sample dataset and the design, training, and testing
processes to obtain two different ANNs for (1) the detection and (2) identification of
a burst. The application stage includes the processing of a potentially continuous
transient head trace into individual time windows to be processed through the two
ANNs and a final stage of verification of the obtained results.
The technique was applied to a numerical example and validated with data obtained
from a laboratory test at the University of Adelaide. In both cases, the technique was
successful at detecting and identifying the occurrence of a burst in a pipeline. In the
numerical application, a sharp burst was considered. The burst detection ANN
accurately classified the analyzed time windows in one of three possible categories
149 |
ADE | Chapter 7 – Fluid Transients and ANN for Burst Detection and Identification
with only a short time lag of 0.016 s to identify transient head traces in the third
category. The burst identification ANN predicted the occurrence of a burst 0.85 m
from the real burst location (an error of 0.085% on the 1,000-m pipeline) with an error
in the prediction of the burst size of 0.17 mm.
For the experimental laboratory application, the burst was modeled as a circular orifice
that did not open instantaneously in order to test the technique with transient head
traces that do not contain perfectly sharp transient waves. In this case, the burst
detection ANN accurately classified most of the time windows. However, there was a
slightly longer time lag in the identification of the first head deviation from normal
conditions. For the burst identification ANN, three different trainings were conducted
to test the robustness of the predictions. Each burst identification ANN predicted
different burst locations and sizes; however, all the errors in the location prediction
were within 1.5% of the total length of the pipeline. The burst size predictions were
less accurate, oscillating between 6% and 16% of the real burst size.
Because this is the first application using fluid transient waves and ANNs, the training
samples were obtained from a numerical model. However, future applications could
use historical data from real systems and consider more general situations in terms of
the burst characteristics and hydraulic systems analyzed. An application of this
technique in a more complex hydraulic system might involve the development of more
robust ANNs, the generation of more training samples, and the placement of more
sensors in the system to collect enough information.
150 |
ADE | Chapter 8
8 Conclusions
Water transmission pipelines constitute a key component of water infrastructure as to
their role in transporting water over long distances. However, their length and the fact
that these pipelines are mostly underground, make their inspection and maintenance
very challenging. Modern, versatile, cost-effective and accurate inspection techniques
for water pipelines are urgently needed. This chapter presents a summary of the
developed work, conclusions and contributions of this thesis including
recommendations for future work.
8.1 Summary of developed work
This PhD research presents the development of a new set of techniques for the active
and passive noninvasive inspection of water pipelines using for the first time custom-
designed Artificial Neural Networks (ANNs) to interpret fluid transient pressure
waves.
The active inspection technique locates and characterizes topological elements and
anomalies in pipelines after the generation of a controlled small magnitude transient
event in a single pipeline. This location and characterization is conducted by
interpreting the reflections that these features induce in the pressure transient wave
using ANNs trained to detect a particular feature. This technique has been expanded
to locate leaks in pipelines where background pressure fluctuations are present by
developing a framework for the training of a set of ANNs using numerical datasets
with the addition of different intensities of noise.
151 |
ADE | Chapter 8 – Conclusions
The passive inspection technique identifies, locates and characterizes abnormal events
occurring in pipelines such as bursts by continuously analyzing the transient pressure
trace obtained from a pipeline without the generation of an artificial transient event.
This ANN-based technique is able to first classify the status of the pipeline as normal
or abnormal, and then predict the location and size of an occurring burst when the
condition of classified as abnormal.
Both techniques (active and passive) proposed in this PhD thesis have been validated
in numerical and experimental pipelines demonstrating the potential of using machine
learning algorithms such as ANNs for the noninvasive inspection of water pipelines.
In contrast to previous research, the techniques included in this thesis are data-driven
in their application because no detailed information from the analyzed pipeline is
required after the training of the ANNs. Accurate, fast and reliable results have been
achieved.
This thesis has opened up the possibility of using deep learning merged with high-
frequency pressure measurements for a fast and reliable condition assessment of
pipelines.
8.2 Research Conclusions
The research work developed in this PhD thesis has demonstrated that Artificial
Neural Networks (ANNs) can be used for the interpretation of fluid transient waves.
Specific conclusions of this work include that:
- Deep artificial neural networks such as 1D convolutional neural networks can
interpret transient pressure traces to predict accurately the location and the
diameters before and after a pipeline junction.
- Deep artificial neural networks such as 1D convolutional neural networks can
recognize the pattern of pressure fluctuations induced by the presence of a leak
in a pipeline to predict accurately its location and size.
- The performance of Artificial Neural Networks in predicting the location of a
change in diameter in a pipeline is enhanced by conducting a time-wise
152 |
ADE | Chapter 8 – Conclusions
downsampling process to the pressure transient signals in comparison to using
raw high-frequency pressure measurements.
- The number of transient pressure samples required for the successful training
of ANNs depend on the anomaly to be detected. If the anomaly produces
important fluctuations in the transient pressure trace (e.g. a change in diameter)
fewer training samples are required in comparison to the samples required to
locate more subtle anomalies (e.g. small leaks).
- Stochastic resonance principles can be used to enhance the performance of an
ANN for the location of leaks in a laboratory based pipeline experiencing
background noise.
- The deployment of stochastic resonance for the location of leaks demonstrated
the existence of an optimum noise intensity to be added to the ANN training
samples. Both the performance during training and when real pressure
transient traces are analyzed need to be considered to determine this optimum
noise intensity.
- A sliding time window analysis using 1D convolutional neural networks can
be used to analyze in real time a pressure transient trace to detect the
occurrence of a burst. The length of this time window needs to be long enough
to capture the transient pressure wave fluctuations and needs to move one data
point at a time to ensure a rapid detection.
- The detection, location and characterization of the occurrence of bursts in
pipelines can be achieved by using two 1D convolutional neural networks
analyzing a continuous transient pressure signal. While one ANN can detect
unusual transient pressure fluctuations, a second ANN can accurately locate
and characterize the size of the burst in a single pipeline.
153 |
ADE | Chapter 8 – Conclusions
8.3 Research Contributions
The overall contribution of this research is the development of novel techniques for
the active and passive inspection of pressurized water pipelines using custom-designed
Artificial Neural Networks (ANNs) to interpret transient pressure waves. Specific
research contributions are as follows:
1. The development of a framework for merging the use of fluid transient waves
and ANNs for the inspection of pipelines (Aim 1). This thesis describes for the
first time that ANNs can be used for the interpretation of transient waves
reflections induced by the presence of different features in a pipeline. This
framework includes methods for i) the generation of multiple numerical
transient pressure traces for the ANN training simulating the presence of the
analyzed feature along the pipeline, ii) the application of a timewise
downsampling for facilitating the ANN training and iii) the selection of an
appropriate ANN architecture between the standard fully connected dense
network and a 1D-convolutional neural network. This research has
demonstrated that the number of numerical transient pressure traces required
for a successful ANN training depends on the feature that needs to be detected.
If the reflections in the transient pressure wave are more subtle (e.g. say due to
the presence of a leak), the number of numerical transient pressure traces
required is larger. It has also been demonstrated that a 1D-convolutional
network is able to predict more accurately the location of a feature (e.g. a
pipeline junction) in comparison with a dense network and has more potential
to be successfully implemented in more complex applications. Finally, the
developed framework indicated the importance of conducting a timewise
downsampling to scale down the time resolution of the transient pressure
traces, which reduced the computational time and enhanced the performance
of the ANN in predicting the location of the features.
2. A methodology for the active inspection of pressurized pipelines has been
developed in this research. This novel technique is able to identify topological
elements such as junctions and to detect, locate, and size anomalies such as
leaks using ANNs as a tool to interpret the resulting transient pressure traces
after the generation of an artificial and controlled transient event (Aim 2.1).
154 |
ADE | Chapter 8 – Conclusions
This methodology uses the framework described above to generate numerical
samples that simulate potential locations and characteristics for a feature and
trains a 1D-convolutional network based only on this information. Once the
ANN has been trained, the methodology can be used by only processing a
different transient pressure trace to obtain a prediction of the feature location
and characteristics. This methodology has been applied to the location and
sizing of a junction and a leak in a numerically modeled pipeline where
satisfactory results were achieved (Aim 4) demonstrating that machine learning
algorithms can be applied to the noninvasive inspection of pipelines.
3. The development of a framework for adapting the active inspection technique
for the location of leaks when the analyzed pipeline is under more realistic
situations such as the presence of background pressure fluctuations (Aim 2.2).
This research has demonstrated that when background pressure fluctuations
are present in pipelines, the performance of the active inspection technique can
be significantly enhanced via deployment of stochastic resonance. When no
noise intensity is included in the dataset or this intensity is very small, the
ANNs trained have an unacceptable performance in their predictions of the
leak location mainly because background pressure fluctuations are not
reproduced by numerical transient flow models and the ANN training is not
successful. By training a set of ANNs using datasets with the addition of
different intensities of Gaussian distributed noise, the distribution of the leak
location predictions narrows around the real leak location demonstrating the
existence of an optimal noise intensity. This research has also indicated that to
select the optimum noise intensity a combined analysis of the distribution of
the predicted leak locations and the RMSE of the training and testing of the
ANNs should be conducted. These results demonstrate for the first time that
stochastic resonance can be applied to the leak detection problem in water
pipelines representing an important stepping-stone in developing a fully
automated methodology for leak detection in pipelines using ANNs to interpret
transient pressure waves.
4. To analyze a potentially continuous transient pressure trace, a methodology for
dividing this signal into short time windows has been developed and
constitutes another research contribution of this PhD thesis. This methodology
155 |
ADE | Chapter 8 – Conclusions
includes the definition of an appropriate length to capture the sudden drop in
pressure from the occurrence of a burst at any location of the pipeline (Aim
3.1). The successive analysis of time windows by moving one data point at a
time allows for the classification of the condition of the pipeline into a normal
or abnormal condition. These transient pressure time windows constitute the
training dataset for the ANNs developed as part of the passive inspection
technique.
5. The development of the passive inspection methodology for the detection,
location and characterization of abnormal events (such as bursts) in pipelines
represents another important contribution of this research (Aim 3.2). This
methodology includes a model development stage and an application stage.
The model development stage includes the generation of a transient pressure
sample dataset and the design, training, and testing processes to obtain two
different ANNs for (1) the detection and (2) identification of a burst. The burst
detection ANN is responsible for classifying a given pressure transient time
window into three possible categories. The burst identification ANN analyzes
the transient pressure time windows that contain the first cycle of reflections
of the burst transient wave along the pipeline to predict the location and the
size of the burst. The application stage includes the processing of a potentially
continuous transient pressure trace, its division into individual time windows
to be processed through the two ANNs and a final stage of verification of the
obtained results. This passive inspection methodology represents the first
application of ANNs to interpret continuously measured transient pressure
traces to detect, locate and characterize the occurrence of bursts and has been
successful in the detection of the occurrence of bursts in numerical and
experimental pipelines.
6. Experimental verifications have been successfully conducted in this research
for the active ANN inspection technique (Chapter 6) and the passive ANN
inspection technique (Chapter 7) developed as part of the PhD thesis. These
experiments have contributed valuable information for the development of
these inspection methodologies and have provided an insight into the potential
of applying machine learning algorithms such as ANNs for the interpretation
156 |
ADE | Chapter 8 – Conclusions
of transient pressure traces while highlighting potential challenges for the
application of these techniques in more complex systems.
8.4 Recommendations for Future Work
Innovation is the driver of research. Digital technologies are rapidly evolving and
supporting almost every aspect of society, including the water industry. The novel
techniques for the inspection of pressurized pipelines developed in this research
constitute an innovative advancement in the knowledge and the application of
noninvasive techniques using ANNs for the interpretation of fluid transient pressure
waves. The combined use of the research knowledge developed for more than 25 years
in fluid transient pressure waves for pipeline inspection and the advantages of artificial
intelligence algorithms have significant potential for contributing to the development
of asset management strategies in water utilities.
The further development of methodologies combining fluid transient waves and
artificial intelligence could result in fully automated techniques that may be able to be
deployed in real pipelines to assess its condition. The monitoring system could be
installed permanently to detect abnormal events without the need of manual analysis
or specific pipeline modelling. A series of future research opportunities have been
identified to potentially improve and expand the methodologies proposed in this
research, including:
1. An exploration of diverse alternatives to create the datasets required for the
ANNs training. The applications presented in this thesis considered the
creation of the training and testing dataset based exclusively on numerical
transient pressure traces. Future applications can consider the combination of
numerical transient pressure traces and historical information available from
the analyzed pipeline. By combining the origin of the samples for the training
dataset, the performance of the ANNs can be made to be more robust and
accurate as real transient pressure traces are included in the training process.
Another potential modification to the ANN training dataset could include the
use of an unsteady flow numerical models including the effects of unsteady
friction and pipeline viscoelasticity to generate the training samples. This
would equip the dataset with samples that more closely resemble measured
transient pressure traces after the first set of wave reflections providing more
157 |
ADE | Chapter 8 – Conclusions
information for the ANN to learn from. Other alternatives to create the ANNs
training dataset for the active inspection technique would include i) the use of
transient pressure traces obtained from different transient wave generation
methods (such as pulse waves or persistent signals), ii) the potential
transformation of the transient pressure signal before the ANN training for both
the active and the passive inspection techniques, iii) the generation of non-
dimensional transient pressure traces to potentially train one ANN able to
detect anomalies or abnormal events in any pipeline. By conducting a non-
dimensional transformation, the trained ANNs could be robust to pipelines
with different dimensions, hydraulic conditions and operational changes.
2. An analysis of different options for the development of fully automated
processes for the active and passive inspection of pipelines. Chapter 5 and 6
summarized the methodology for the active inspection of pipelines after the
generation of a transient event where a selected segment of the measured
pressure signal is processed through the ANN (or set of ANNs) to obtain a
prediction of the location and characteristics of the anomaly. A potential
improvement to this approach would include the application of the sliding time
window concept proposed for the passive inspection of pipelines in Chapter 7
to identify the transient pressure time window of interest by analyzing the
continuous transient pressure signal obtained after the test is conducted in the
pipeline. This approach would provide the technique with tools to conduct the
active inspection of the pipeline automatically while exposing the ANNs to
more information about the analyzed pipeline.
3. The application of the active and passive inspection techniques developed in
this research to other anomalies or other abnormal events expected in water
pipelines and to more complex pipeline topologies. Although the principles of
the methodologies presented in this thesis are applicable to any potential
anomaly or abnormal event in a pipeline (as long as those anomalies are
discernable in a transient pressure trace), applications have been presented only
for the location of leaks and bursts. Future applications of this research would
include the detection of blockages, pipe wall deteriorated sections, partially
closed valves, abnormal surges in water demand or sudden valve closures.
Similarly, these methodologies can be expanded to the analysis of more
158 |
ADE | Chapter 8 – Conclusions
complex topologies including pipelines in series or looped networks using
information from multiple pressure measurement points or by combining the
use of ANNs with other existing transient based pipeline inspection techniques.
4. The enhancement of the proposed the passive inspection methodology by
analyzing detection results using standard metrics such as Receiver Operating
Characteristic (ROC) curves and Confusion Matrices (Myrans et al. 2018).
Using these tools will provide a better understanding of burst detection results
in terms of trade-offs between true detection rate and false positive and false
negative rates. This analysis can point out areas of improvement in different
steps of the methodology and can also serve as comparison metrics between
the proposed methodology and existing techniques.
5. Considering that the both the active and passive techniques proposed in this
research have proven effective in numerical and experimental applications,
future research would focus in comparing these techniques with existing
methodologies for the location of leaks and bursts in pipelines. This
comparison should be conducted including aspects such as robustness to noise,
detection accuracy and computational time required.
6. Although this thesis includes experimental validation of the active and passive
inspection techniques, future work should be focused on more extensive
laboratory and field verification of these techniques. This verification could
include different aspects such as different leak/burst sizes (to determine a range
of detectable anomaly sizes for a pipeline) and different preprocessing
procedures such as signal denoising or signal transformation. In addition, a
wider exploration of ANN configurations and training samples should be
conducted to determine if better results can be achieved with alternative ANN
architectures with different rates of time-wise downsampling.
159 |
ADE | Characterization of transient pressure traces due to the
effects of different anomalies and features in water
pipelines
J Bohorquez, A Simpson, M Lambert
School of Civil, Environmental and Mining Engineering,
The University of Adelaide, Australia
ABSTRACT
Different features or anomalies can be present in pipelines. This paper presents the effect
that different features have on the resulting pressure trace after a transient event when the
location of the anomaly changes along the pipeline and when the transient generation and
measurement point is changing. It has been found that locating the transient generation
and measurement point in the upstream half of the pipeline, close to the middle point is
desirable because it prevents larger pressures in longer sections of the pipeline while
keeping the reflections from the features in the first pressure plateau. An analysis of the
effect of having a linear valve closure for the transient generation instead of an
instantaneous one is also presented, showing that real field conditions can make the
identification of features more difficult. In addition, the effect of the symmetrical
location of features on the pressure trace has been studied, showing that two different
anomalies or features located on alternative sides of the generation point, induce the
same effect in the first pressure plateau, complicating its identification.
1. INTRODUCTION
In comparison with other civil engineering infrastructure such as bridges or buildings,
inspecting the condition of pipelines is more challenging because they are usually
underground. Several different non-invasive approaches have been proposed to locate
anomalies or features in pipelines. Acoustic (1), ground penetrating radar (2) and
electromagnetic (3) techniques have been used to locate mostly leaks. However, during
the last 25 years, a new set of techniques has been developed using fluid transient based
methods. For these approaches, a small controlled pressure transient signal is generated
in the system and the interpretation of its pressure trace response allows the location of
different features.
The evolution of these techniques has been focused on leak location (4-6) yet attention
has expanded to other features with applications for internal deterioration sections (7-9),
locating blockages (10), bursts (4), air pockets (11), and determining the status of closed
or partially closed valves (12). In addition, the different techniques based on transient
pressure waves can be classified into four categories (13): inverse transient methods (14),
transient reflection methods or time reflectometry methods (15), system response
methods (16) and transient damping methods (17).
© BHR Group 2018 Pressure Surges 13 151
208 |
ADE | Despite the development of transient based techniques, and the existing knowledge of the
effect of the presence of features or anomalies in the pressure traces, no joint and
comparative analysis on the differences between the common analysed features has been
previously conducted. These features include the presence of deteriorated sections, leaks,
blockages, junctions, plastic pipe replacement segments and air pockets. In addition, the
influence in the pressure traces by the transient generation (and measurement point)
location and its interaction with the features has not been analysed either.
This paper presents a description of the system and the features considered, followed by
an analysis of the influence of the transient generation location in the pressure trace of an
intact pipeline (one without anomalies). In order to differentiate the effect of each
feature, pressure traces for all of them are presented when the transient event is generated
at the end of the pipeline, highlighting their differences. In addition, an analysis when the
position of the generator and the feature is moved along the pipeline is included by
showing important remarks for each feature. The effects on pressure traces are studied
for two additional situations: a linear valve closure for the transient generation instead of
an instantaneous one and the presence of two features that are the same distance from the
generation point but in opposite directions along the pipeline.
For all the results presented in this paper, the pressure traces have been obtained from
numerical simulations using the MOC (18) considering steady state friction and
neglecting the effect of pipe wall viscoelasticity in the plastic pipes that were modelled.
2. PIPELINE CONFIGURATIONS
The system configuration selected to develop the proposed characterization is shown in
Figure 1. In general, it was considered a single pipeline with a total length
connected to a reservoir in the upstream end and to a dead end at the downstream end.
The position of the feature is defined by the distance from the reservoir to its location
and the position of the side discharge valve that represents the transient
generator (denoted G/M in the figure since it corresponds to the generation and
measurement point) is represented by .
Figure 1. Pipeline configuration.
Details of the selected characteristics for the described pipeline are shown in Table 1.
This set of parameters was fixed for all the developed simulations. The position of the
feature and the generator are not presented in this section because those were changed
along the pipeline in a number of different simulations. In addition, the different
characteristics associated with the analysed features will be presented in the following
section.
152 © BHR Group 2018 Pressure Surges 13
209 |
ADE | Table 1. Pipeline characteristics.
Characteristic Value
Total Length (m) 1000
Steady state head at reservoir (m) 40
Internal Pipe diameter (mm) 300
Wall Thickness (mm) 6
Youngs Modulus (GPa) 210
Steady state velocity (m/s) 0.045
Flow at side valve (L/s) 3.18
Steady state Reynolds number (-) 11842
Wave speed (m/s) 1191
The base flow through the pipe was set to zero since the downstream boundary condition
is a dead end (which in a real field test represents a closed valve that allows the isolation
of the studied pipeline). Accordingly, the only flow through the pipeline is the flow
through the side valve (the transient generator) that is closed instantaneously to induce
the transient event (except for the results in Section 7). The considered pipeline is
assumed to be made of steel with an intact wall thickness of 6 mm and for the set
parameters; the intact pipeline has a wave speed of 1191 m/s.
3. FEATURE CHARACTERISTICS
Six different features have been considered in the analysis: a leak, a blocked segment, a
deteriorated segment, a junction, a pipe repair with plastic (PVC) section and an air
pocket. Each one of these features will affect the pressure trace in a different way given
their characteristics and their interaction with the generated transient wave. Figure 2
presents a diagram that summarizes what each feature represents (and will be used in the
rest of the paper) and their main characteristics.
Figure 2. Features characteristics.
© BHR Group 2018 Pressure Surges 13 153
210 |
ADE | The presence of a leak in the pipeline is characterized according to the flow of the leak
and its equivalent diameter (modelling the leak as a circular orifice with a fixed diameter,
a discharge coefficient of 0.6 and atmospheric pressure on the downstream side of the
orifice). Considering these parameters, the leak flow represents a 37% of the flow that
goes through the side discharge valve in the pipeline before the valve closure, which
corresponds to a flow rate of 1.17 L/s.
The existence of a blocked or deteriorated section is characterized by the length, the
resulting internal diameter and the wall thickness of the section. These parameters will
affect its wave speed. Even though the blocked section is modelled as a section with a
significantly lower internal diameter, the wall thickness was not modified since the
existence of a blockage does not affect the interaction between the pressure and the pipe
wall.
The junction feature represents a situation in which the analysed pipeline is connected to
another pipeline which at its downstream end also has a dead end (not shown in the
figure). Different lengths for the junction are considered but due to brevity only the
results for a length of 700 m are presented. The pipeline that joins at the junction is
assumed to have the same diameter and characteristics of the main pipeline (in terms of
material and wall thickness).
The fifth considered feature is a plastic pipe segment with the same inner diameter of the
steel pipe segment, but with a slightly greater wall thickness (selected from an Australian
plastic pipe supplier catalogue) of 7 mm. The length of the repaired segment was set in
12 m to represent two of the commercially available pipe segments and the Youngs
Modulus was set in 3.07 GPa to represent, in the wave speed computation, the change in
material. Finally, the sixth and last feature is an air pocket which is only characterized by
its volume, set in 0.005 m3 to represent a small air pocket.
When the different features were incorporated to the MOC, the leak, junction and air
pocket were modelled as inner boundary conditions to the model; whereas, the blockage,
deteriorated section and the plastic repaired where modelled as a new section with the
indicated characteristics. For these latter cases, the length of the downstream segment of
the intact pipe was modified to preserve the total length of the pipe.
4. TRANSIENT GENERATION ALONG AN INTACT PIPELINE
The first set of conducted simulations were developed for an intact pipe (with no
presence of the described features) by changing the location of the generation and
measurement point to the five locations shown in Figure 3. The simulation time was set
in 4.0 s, which allows the analysis of the first seconds considering the total length
of the pipe and the intact wave speed.
Figure 3. Generation/Measurement location positions considered.
154 © BHR Group 2018 Pressure Surges 13
211 |
ADE | These locations were selected to examine the effect of generating a transient event at
different points of the pipeline. The distances presented in the lower part of the figure
correspond to the distance measured from the reservoir to the potential
generation/measurement point. Three different positions were considered before the
midpoint of the pipeline and two positions were set in the downstream half of the
pipeline. The resulting pressure traces for the five locations are presented in Figure 4.
Figure 4. Pressure trace for an intact pipe for different generation/measurement
points.
For all the locations, the transient event generated was associated with the same flow
through the side valve that was assumed to be closed instantaneously. However, the
initial head rise differs if the generation point is away from the downstream end of the
pipeline (positions 1, 2, 3 and 4) or at the end of the pipeline (position 5). If the valve is
located away from the downstream end of the pipeline, the change of flow (or change in
© BHR Group 2018 Pressure Surges 13 155
212 |
ADE | velocity) that induces the head rise propagates both upstream and downstream of the
valve, inducing a lower head rise when compared to the generation of the transient at the
end of the pipe. In this last case, the change in flow propagates only in the upstream
direction thereby generating a larger pressure head rise.
When the resulting pressure traces for generation and measurement point locations 1, 2
and 3 (250, 400 and 480 m from the reservoir respectively) are analysed, is evident that
the Joukowsky head rise is not exceeded. The differences between the traces correspond
to the arrival of the reflections from the boundary conditions (either the reservoir or the
dead end) to the measurement point at different times. For example, when the generation
and measurement point is located 250 m downstream from the reservoir, the reflection
from this boundary condition (that causes a pressure release) arrives earlier than the same
reflection when the generation and measurement point is located at 400 m.
In general, as long as the generation and measurement point is located in the upstream
half of the pipeline, the maximum pressure recorded would correspond to the initial
Joukowsky head rise given that the reflection from the reservoir (that allows the releasing
of pressure) arrives to the measurement point before the reflection from the dead end. To
the contrary, if the generation point is located in the downstream half of the pipeline (as
is shown when the generation point is located 830 m downstream of the reservoir) the
maximum pressure reaches two times the initial head rise given that the reflection from
the dead end arrives before the pressure release from the reservoir.
When the generation point is located in the upstream half of the pipe, the maximum
recorded pressure is the Joukowsky head rise; however, part of the pipe is still
experiencing a double head rise due to the reflection from the dead end. In general, the
closer the generation point is to the reservoir, the shorter the segment of the pipeline that
will be exposed to a larger pressure. In contrast, locating the generation point near the
dead end will induce the larger pressure in a longer segment of the pipeline.
Following this analysis, locating a side discharge valve for generating transient events in
the upstream half of the pipeline will be beneficial because the length of the pipe that is
exposed to a larger pressure will be minimized for a given initial flow through the side
discharge valve.
5. TRANSIENT GENERATION AT THE END OF A PIPELINE WITH
FEATURES
A second set of numerical simulations has been conducted to characterize and
differentiate the effect of each feature in the pressure trace, when the transient event is
generated at the dead end of the pipeline. For these simulations, the location of the
different features ( in Figure 1) was set in 300 m and the simulation time was
4.0 s (time that again covers the period in all the considered cases). The resulting
pressure traces are presented in Figure 5 where a diagram in the lower left corner of each
plot indicates the analysed feature.
The resulting pressure traces from Figure 5 allows some features to be detected as they
trigger subtle effects in the pressure trace and others that have more complex effects in
the traces. The presence of a leak and a deteriorated section induce slight changes in the
pressure while features such as a blockage, a junction, a plastic segment and an air pocket
generate more prominent perturbations of the pressure.
156 © BHR Group 2018 Pressure Surges 13
213 |
ADE | Figure 5. Effect in pressure traces when features are located 300 m downstream the
reservoir and the generation point is at the end of the pipeline.
First, the leak is characterized with a slight drop of pressure (see Figure 5-a) that is
always present in the first pressure rise plateau, as long as the generation point is at the
end of the pipeline. Consecutive perturbations in comparison to the pressure trace
corresponding to an intact pipe can also be seen in the rest of the simulation time as drops
or rises from the interaction of the leak with the wave reflected in the boundary
conditions. In general, even with the presence of the leak, the behaviour of the pressure
equivalent to an intact pipe is still visible.
The second analysed feature is a 10 m length blockage. In the presence of this feature,
the pressure trace shows a spike in the first head rise plateau (see Figure 5b) after the
valve closure and after some repetitive reflections inside the blocked section, the pressure
© BHR Group 2018 Pressure Surges 13 157
214 |
ADE | returns to the initial pressure head rise. The duration and magnitude of the spike are
directly related to the length and diameter of the blocked section. Following this first
pressure spike, multiple reflections on the blocked section are visible as smaller spikes
(when a positive wave is travelling in the pipeline) or small drops (as the one visible at
3.0 s) when a negative wave is moving along the pipeline. Although the effect of the
blockage is more pronounced than the effect of a leak, the main transient wave is still
visible in the trace.
Third, a 10 m deteriorated section is analysed. In general, this feature induces a slight
drop and recovery of the pressure in a short period of time (see Figure 5-c). How short
the drop is in time depends on the length of the deteriorated section and the magnitude of
the drop is related to the difference in the wave speed of the two sections. However, a
deterioration is understood as the loss of wall thickness or the cement mortar lining,
depending on the case, and would never induce significantly big changes in wave speed.
As the effect of the deterioration is slight, the intact pipe pressure trace is still visible.
However, multiple reflections are visible after the first pressure rise that could be
confused with the presence of other deteriorated sections or small blockages.
The effect of the presence of a junction in a pipeline is shown in Figure 5-d. Unlike the
presence of a leak, blockage or deterioration, the presence of a junction induces a
complete perturbation of the pressure trace, which is more difficult to analyse. In
addition, the response of the pressure trace is highly sensitive to the diameter, the length
and the boundary condition at the end of the junction pipe. In general, in the first pressure
rise plateau, the first visible effect of a junction is a drop in pressure followed by the
reflection of the first wave at the dead end of the junction (for this junction location).
However, that plateau is interrupted by multiple reflections and is no longer visible as in
the previous features and the rest of the trace becomes even more complex to interpret.
Although the characterization of the junction is difficult only by using the pressure trace,
an interesting fact is visible in the trace. The maximum pressure during the simulation
time is larger than the initial Joukowsky head rise, showing that more complex
topologies not always guarantee that the transient pressure dissipates faster, as it has been
pointed out before (19).
The presence of a 12 m plastic pipe repaired segment was considered as the fifth feature
(see Figure 5-e). It was included because replacing segments of damaged, deteriorated
pipe with plastic new pipes is common in water distribution systems and its effect is still
not well understood. The main difference in the transient behaviour of these pipes is due
to its low hydraulic impedance given the lower wave speed. In fact, in the first pressure
rise plateau the effect of the presence of a plastic pipe segment is a significant drop in
pressure resulting from the reflection on this low impedance section. However, unlike the
presence of a deterioration, the pressure drop is greater and is always followed by a
stepped recovery of the initial pressure rise (which is related to the accumulation of
pressure in the plastic pipe). In this sense, a plastic segment has a similar effect to the
blockage but with the opposite effect, instead of inducing a pressure increase, it triggers a
pressure decrease. In addition, as is visible in the figure, the rest of the pressure trace,
after the first plateau is affected by the existence of this plastic segment, even though its
length is short. In general, pressures in the pipe are lower in comparison to that of the
intact pipe pressure trace. This behaviour corresponds to previous studies reported on the
role of plastic pipes in metallic systems (20).
The last feature included in Figure 5 is an air pocket. Presence of small air pockets have
been reported as a source of error for condition assessment techniques that use transient
158 © BHR Group 2018 Pressure Surges 13
215 |
ADE | waves (11) and the results in this paper confirm this. If the pressure trace (see Figure 5-f)
corresponding to the presence of an air pocket is compared to the one corresponding to a
plastic pipe segment (see Figure 5-e), some similarities are evident. First, in the initial
pressure rise plateau, both features induce a significant drop in pressure followed by a
recovery of the initial pressure head rise. Second, reflections after the first period of the
pipe have common elements as a pressure spike around 2.5 s and 3.2 s. In addition, the
presence of an air pocket reduces the wave speed of the fluid in the pipeline, modifying
the general patterns of the pressure trace when compared to the one corresponding to an
intact pipe.
In general, all the considered features induce different effects in the pressure trace when
the transient is generated at the end of the pipeline and the first pressure rise plateau is
analysed. Only the air pocket showed strong similarities to the plastic repaired pipe
segment. This characterization, despite being qualitative instead of quantitative, is
important in a process of developing techniques that can eventually determine the
location and characteristic of features in water pipelines based on the analysis of the first
pressure plateaus or using the whole pressure trace.
6. TRANSIENT GENERATION AND FEATURE POSITION ALONG THE
PIPELINE
The third set of simulations is oriented to combine the previous two considerations:
moving the transient generation (and measurement) point along the pipeline and
analysing the effect of the six selected features in different positions. In total, 96 different
pressure traces have been obtained by moving the generation point over the five positions
described in Figure 3 and locating the six features in the four remaining positions
(excluding the dead end). Key findings are identified by analysing all the pressure traces
but for brevity only some of the plots are shown, the ones that allow the illustration of the
different observations. The purpose of this analysis is not to select a transient generation
(and measurement) point along the pipeline that is the optimum location in order to
identify each feature, but rather to highlight different trends and behaviours of the
pressure trace depending on various selected positions of the transient generation point
and the possible location of a feature.
6.1. Leaks
Figure 6 presents three pressure traces corresponding to three different positions of a leak
along a pipeline where the transient event is generated 480 m downstream from the
reservoir. In order to facilitate the observation, a diagram showing the position of the
leak and the generator is presented.
If a time reflectometry analysis is conducted, identifying the different features when their
reflections are observed in the first pressure head rise facilitates the analysis because it
decreases the probability of recording reflections of other elements or the loss of
information due to dampening of the transient wave. In this sense, locating the generation
point close to any of the boundary conditions would not be desirable given that the
reflection either from the reservoir or from the dead end would modify the first pressure
head rise.
© BHR Group 2018 Pressure Surges 13 159
216 |
ADE | Figure 6. Pressure traces when Generation/Measurement point is at 480 m and a
leak is at three different positions: 250, 400 and 830 m.
Accordingly, positioning the generation point close to the midpoint of the pipe is
beneficial because almost any reflection from a feature reaches the measurement point
before the reflection from a boundary condition. An example of this is presented in
Figure 6. By locating the generation point at 480 m, the pressure drop corresponding to
these features is observed in all cases before the reflection from the reservoir (which
causes the prominent pressure drop). To the contrary, if the generation point is located at
any other internal point of the pipeline, the reflection from a leak could reach the
measurement point after the first prominent reflection, or even at the same time,
preventing easy determination of its location.
It is important to note that choosing the midpoint of the pipe exactly as a generation point
is not desirable either. Firstly, the reflection from the boundary conditions would reach
the measurement point exactly at the same time (and therefore may cancel each other
out) and secondly, by having that setup it would not be possible to establish the exact
location of the feature (there would be two possible positions, one on each side of the
measurement point). Therefore, if the midpoint of the pipe is chosen as the generation
and measurement point, two or more measurement points are required to obtain the exact
location of a feature. This observation is analysed in more detail in Section 8.
However, in a field application the generation point should perhaps not be as close to the
midpoint of the pipeline as it is in this numerical example to prevent a similar effect as
having the generation point exactly at the mid-point of the pipe. In this real application,
the valve closure is non-instantaneous (as is further analysed in Section 7) and therefore
the pressure rise is not exactly sharp. Thus if the generation point is too close to the mid-
point the final pressure trace could have reflections from the boundary conditions
occurring at the same time, making the analysis more difficult.
6.2. Blockages
The second feature is a 10 m pipe section that is partially blocked (according to the
characteristics shown in Figure 2). For this feature, results (Figure 7) are presented again
for the case when the generation and measurement point are located 480 m downstream
the reservoir and the blockage is 250, 400 and 830 m respectively downstream this
boundary condition.
160 © BHR Group 2018 Pressure Surges 13
217 |
ADE | Figure 7. Pressure traces when Generation/Measurement point is at 480 m and a
blockage is at three different positions: 250, 400 and 830 m.
In this figure, it can be seen that by locating the generation and measurement point near
the midpoint of the pipe, the pressure spike corresponding to the blockage is always
visible in the first pressure rise after the valve closure. However, unlike the leak pressure
trace, the presence of the blockage also affects the reflections from the boundary
conditions.
For the first two cases, when the blockage is between the reservoir and the generation
point, the reflection from the reservoir interacts first with the blockage before reaching
the measurement point, inducing the stepped drop followed by an instantaneous rise in
pressure that is visible after the first pressure spike. This is visible in the pressure trace at
1 s in Figure 7-a and Figure 7-b. To the contrary, when the reflection from the dead end
goes through the blockage before reaching the measurement point (which happens when
the blockage is between the generation point and the dead end), a stepped rise in pressure
is visible after an instantaneous drop due to the presence of the reservoir (see Figure 7-c).
Under these circumstances, when the feature significantly affects the reflections from the
boundary conditions, the reflections could be confused with other features. For instance,
in Figure 7, given that the pressure drop corresponding to the boundary conditions is
affected, it looks similar to the effect of the presence of an air pocket, or even a severe
deterioration. Therefore, even while locating the generation point close to the midpoint
of the pipeline can be convenient for detecting features in the first pressure rise; it can
also induce misidentifications given the interaction of the features with the boundary
conditions.
6.3. Deteriorations
Deteriorated sections are always of interest in the pipe condition assessment area because
these sections could be more prone to suffer bursts or develop leakages. Thus, it has been
analysed in depth by different researchers. As it was described above, the effect of a
deterioration in a pressure trace tends to be subtle given that the reflections come from a
slight change in the wave speed of the pipeline. Figure 8 presents the resulting pressure
trace when the 10 m deteriorated section is located at 250, 400 and 830 m and the
transient event is generated at the end of the pipeline.
© BHR Group 2018 Pressure Surges 13 161
218 |
ADE | Figure 8. Pressure traces when Generation/Measurement point is at the dead end
and a 10 m deteriorated section is at three different positions: 250, 400 and 830 m.
By observing the resulting pressure traces, is evident that for all the considered locations,
the reflection from the deterioration is visible in the first period of the transient event
(which corresponds to 1.68 s in Figure 8-a). This fact is convenient because it eliminates
the problems observed in the previous two features. First, no matter the location of the
feature, by generating and measuring at the end of the pipeline all the features in the
pipeline would be visible in the first head rise. Second, there is no possibility of
confusing the reflections from the boundary conditions with other features because the
only boundary condition visible in the trace would be the reservoir, visible as a strong
drop in pressure.
Considering this, locating the generation point at the end of the pipeline would be the
most desirable position. However, this setup is usually difficult to accomplish in
pipelines in the field. In general, a valve can be closed at the end of a pipeline to
guarantee the boundary condition but the available locations to install the side discharge
valve for inducing the controlled transient event are usually restricted to say air valves (in
case of above ground pipelines) or hydrants (for urban water pipelines) and those do not
always correspond to the end of the pipe.
In addition, as it was analysed in Section 4, when the transient is generated in the
downstream half of the pipe, a greater proportion of the pipe experiences larger pressures
(corresponding to two times the Joukowsky head rise of a closure in an interior point)
given the reflection at the dead end. If the transient event is generated at the end of the
pipe, the initial pressure rise will be two times the head rise when the event is generated
in an interior point and therefore the entire pipe will be under the effect of larger
pressures. Moreover, if the feature induces spikes or increases in the pressure trace, the
pipeline could be exposed to even larger pressures.
6.4. Junctions
Out of all of the analysed features in this paper, the junction is the one that affects the
pressure trace more prominently. As it was stated in the previous section, its presence is
characterized by a drop in pressure (and its magnitude depends on the diameter of the
junction) followed by multiple reflections that are difficult to interpret. To broaden this
observation, Figure 9 shows the pressure traces of three different junctions when the
generation point is located 250 m downstream the reservoir.
162 © BHR Group 2018 Pressure Surges 13
219 |
ADE | Figure 9. Pressure traces when Generation/Measurement point is at 250 m and a
blockage is at three different positions: 400, 480 and 830 m.
As it can be seen in this figure, the pattern corresponding to the intact pipe, when the
generation is in that position (shown in Figure 4), is affected completely and is no longer
visible, except for the first pressure rise. The pressure drop corresponding to the
reflection of the first wave in the junction is visible no matter the location of the junction.
However, for a junction located further than 250 m (distance to the reservoir) from the
generation point, the drop is not visible in that first rise.
When the junction is located 480 m downstream the reservoir, the pressure drop arrives
almost at the same time as the reflection from the reservoir. In the field, this drop would
have been hidden by the fact that the valve closures are never instantaneous (as it is
analysed in Section 7) making it more difficult to identify this feature. In general, further
analyses are necessary in order to characterize the presence of a junction and on how to
couple this characterization with a condition assessment technique.
6.5. Repaired plastic pipe segments
The combination of different materials in pipes for water systems is a common practice.
Plastic segments of pipes are often used to replace old metallic sections after inspections
or bursts. If a plastic segment is installed and a transient event is generated, the
prominent change in wave speed alters the pressure trace. Figure 10 presents the pressure
trace when a 12 m plastic segment is installed, and the transient event is generated close
to the dead end, at 830 m.
When the pressure traces shown in this figure are compared to the effect observed when
the generation is at the end of the pipeline (Figure 5), one interesting observation stands
out. Even though there is only one plastic segment, the three traces show at least two
patterns corresponding to a plastic segment (a pressure drop followed by a stepped
recovery of the pressure). These multiple effects are visible due to the proximity between
the generation valve and the dead end; all the reflections that arrive at the generation
point coming from upstream, reflect again at the dead end in a short time and are visible
again at the measurement point. In addition, the initial transient wave that propagates
downstream and reflects at the dead end also interacts with the plastic segment sending
additional reflections to the measurement point.
© BHR Group 2018 Pressure Surges 13 163
220 |
ADE | Figure 10. Pressure traces when Generation/Measurement point is at 830 m and a
plastic segment is at three different positions: 250, 400 and 480 m.
For this generation location in particular, the multiple reflections look like successive
plastic segments that are separated only by a short distance (no matter the real location of
the plastic segment); but, as it is shown, only one plastic segment really exists. This
potential problem can be solved by calculating the separation in time between the
repetitive reflections. If they correspond to the feature, they would be equally spread in
time by the time that takes the wave to travel to the dead end and return to the
measurement point. However, this could over complicate the analysis and a better
alternative, if possible, would be choosing a different measurement and generation point.
6.6. Air Pockets
The last feature analysed is an air pocket. Figure 11 shows the pressure trace when the
transient event is generated exactly at the dead end of the pipe (to facilitate the
observation) and the air pocket has three different locations. In general, the presence of
air pockets pose a significant difficulty to the condition assessment task because,
depending on the air volume, they can modify the pressure trace and resemble other
features such as deteriorations or plastic segments.
This figure presents three different locations for the air pocket in which the last one is
close to the dead end. As it can be seen in the figure, when the feature is close to the dead
end (Figure 11-c), multiple reflections are observed in the first pressure plateau. This
could be confused with the existence of several features when there is only one present.
The difference with the highlighted observation shown for the plastic segment (in the
previous sub section) is that the multiple reflections respond to the position of the
features and not to the position of the generation point.
In general, every time that a feature is located in downstream half of the pipe there would
be more than one reflection of the feature visible in the measurement point before the
arrival of the reservoir reflection, but the closer the feature to the dead end, more severe
and more evident that effect is. Unlike the case in which the multiple reflections are
present due to the location of the generation point, the location of a feature cannot be
changed, and any pipe condition assessment technique must consider this possible
phenomenon in order to get valid and accurate results.
164 © BHR Group 2018 Pressure Surges 13
221 |
ADE | Figure 11. Pressure traces when Generation/Measurement point is at the dead end
and an air pocket is located at three different positions: 250, 400 and 830 m.
7. EFFECT OF VALVE CLOSURE
The previous analysis has resulted in different remarks about the position of the
generation and measurement point and the locations of different features. This
characterization is important for defining the different effects in the pressure trace.
However, the conditions of the considered transient event are still idealized in previous
sections. One of the assumptions is that the closure of the valve is instantaneous in order
to get a sharp transient wave travelling through the pipe and this condition is hard to
accomplish in the field and laboratory tests. Figure 12 presents the effect of closing the
side discharge valve to obtain a linear change in velocity in 0.1 s. The figure includes the
pressure trace of the instantaneous closure as a dashed line to facilitate the comparison
between the two cases: the presence of a deterioration and a blockage at 250 m when the
generation and measurement point is at 400 m.
Figure 12. Effect of a non-sharp valve closure in pressure trace when the generation
point is at 400 m and a) a 10 m deterioration, b) a 10 m blockage is at 250 m.
Once the valve closure is no longer considered instantaneous, both pressure traces are
significantly affected. For a deterioration (Figure 12-a) the sharp drop followed by a
pressure recovery (see Figure 8) is no longer visible and is replaced by a slight drop in
© BHR Group 2018 Pressure Surges 13 165
222 |
ADE | pressure that is harder to identify. In addition, the time that it takes for the pressure to go
back to the initial head rise (which corresponds to the reflection from the transient wave
when it moves from the deteriorated section to the intact pipe) is higher, again because
the travelling wave is not sharp.
For the blockage, a similar effect is observed. Even though the general effect of the
blockage is still the same (an increase in the pressure), the spike in pressure is softened
into a gradual increase of pressure when a linear closure is considered. Once again, the
time of the recovery of the initial pressure rise is higher when the valve is closed linearly
than the time when the valve is shut instantaneously. This fact is important because it
could affect the determination of the affected length (with deterioration or blockage) if
this computation is only done by time reflectometry: the feature length would be
overestimated when the valve is closed in 0.1 s.
In addition, both traces show that the pressure response of the pipe is shifted in time now
that the increase in pressure is linear over the 0.1 s. Given this, any technique that
requires the use of the intact pipe pressure trace (as a reference) needs to take into
account the real valve closure pattern.
8. EFFECT OF SYMMETRY IN FEATURES LOCATION
The effect on the pressure trace given the presence of a feature is characterised by the
reflection of the initial transient wave on any singularity in a pipeline. This principle is
used by time reflectometry techniques to identify the location of features by interpreting
the reflections shown in the measurement point in the first seconds, given that this
time allows the recording of a reflection coming from any point of the pipe. However, if
the generation and measurement point is at any interior point of the pipeline, two features
located at the same distance but in opposite directions from the transient generator
location (one upstream and the second downstream) would show the same reflections at
the beginning of the trace. To illustrate this, Figure 13 presents this situation for two of
the analysed features, a leak and a junction.
Figure 13. Effect of symmetry in features positions. The generation point is located
at 480 m and the feature is located 80 m in either direction. a) Leak and b) Junction.
In both cases, the feature was located 80 m upstream and downstream of the generation
and measurement point (400 m and 560 m from the reservoir) which is located 480 m
from the reservoir. Given that the generation point is located close to the midpoint of the
pipe, the initial wave travels across the pipe and comes back to the measurement point in
166 © BHR Group 2018 Pressure Surges 13
223 |
ADE | approximately seconds and therefore only during this time the reflections from the
two features are the same. For both cases, the first reflection arrives at the same time to
the measurement point. For the leak, the first pressure drop is evident in Figure 13-a to
facilitate the understanding. For the junction, the pressure drop is easier to identify in
Figure 13-b. However, if the rest of the trace is analysed, some differences stand out. In
the leak trace (Figure 13-a) the following reflections are opposite, given that the leak is
interacting with opposite boundary conditions each time. For instance, after 2.5 seconds,
before and after the prominent reflection from the reservoir, when the leak is upstream
the measurement point, a pressure drop is observed while for the case in which the leak is
downstream the generation point, a small increase in pressure is visible. Nonetheless, the
differences are still difficult to identify given that the leak effect on the pressure trace is
subtle.
When the same situation is analysed for the junction (Figure 13-b), differences are harder
to identify. The first reflection from the reservoir (visible at 1.0 s) is reduced when the
junction is closer to the reservoir (upstream the measurement point) and enlarged when is
downstream this point. In general, both traces after the first seconds are significantly
different and for the junction in the upstream half of the pipe, the maximum recorded
pressure is larger than the second case.
If a technique for identifying and locating features in pipelines that only uses the first
pressure rise plateau is selected, a single measurement point is not enough to determine
the exact location of the feature. Multiple measurement points would have to be used
given that the reflections to those additional points would not reach at the same time and
by correlating the arrival times at all the measurement points, the exact location of the
feature can be found.
The symmetry in the location of different features can also induce different effects on the
resulting pressure traces. For example, if two leaks are present in a pipeline and the
generation and measurement point is in the middle of them, the recorded pressure drop
would have the same magnitude as if a single leak but with double size was present. For
this case, the individual effects of the two features overlap to magnify the effect. In
contrast, the presence of two features with opposite individual effects (such as a
deteriorated section and a section with a thicker pipe wall thickness due to a different
pipe class) could also overlap but the final effect on the pressure trace could be
diminished or even cancel out.
9. CONCLUSIONS
A characterization and differentiation of the resulting pressure traces in the presence of
different features in a pipeline have been presented. Even though the independent effects
of some of the analysed features have been known in the literature in the past, a
comprehensive overview and analysis of different features and their interaction with the
transient generation and measurement point have been presented in this paper.
Even when some of the features induce similar effects (either increase or drop in
pressure), each one has a particular and distinctive signature. However, the resulting
pressure trace depends on the location of the side discharge valve that generates the
transient event and its interaction with the feature. Locating the generation point at the
end of the pipeline facilitates the analysis of the pressure traces but is a configuration that
is hard to achieve in the field.
© BHR Group 2018 Pressure Surges 13 167
224 |
ADE | E-proceedings of the 38th IAHR World Congress
September 1-6, 2019, Panama City, Panama
2 BACKGROUND
Different methods using transient waves have been proposed for the detection and characterization of
leaks. Liggett and Chen (1994)were the first to use transient signals with this objective. Using inverse transient
methods, a methodology that compares a forward model data with the results of a numerical model, a WDS
was calibrated to identify friction factors, leaks and unauthorized uses. Although successful, this approach is
computationally expensive and requires a detailed numerical model. Later, Lee et al. (2007)conducted a series
of analyses focused on understanding the limitations of using time reflectometry as a technique in laboratory
and field applications. This method is popular due to its simplicity; however, achieving an automatic system that
identifies leak reflected signals, and developing a general method that locates leaks independently from the
measurement and generation points and that manages the influence of the wave speed can be highly
challenging. This mainly again is due to the differences in the head traces from pipelines with different
dimensions and the requirements of knowing these beforehand.To overcome this, some authors have proposed
the use of non-dimensional equations as part of other existing techniques.
Wang et al. (2002) analyzed the transient signal in a pipeline by finding an equation that described the
damping rate of the head according to the characteristics of a leak. As part of these equations, different
dimensionless quantities were defined and incorporated into the unsteady flow equations including a non-
dimensional head, flow, time and leak location. A similar approach is followed in this paper in terms of proposing
non-dimensional quantities but its application is direct to a dimensional head trace obtained either numerically
or in an experiment.
3 SYSTEM CONFIGURATIONAND STEADY STATE MODELING
The system configuration selected to analyze and develop the non-dimensional characterization is
presented in Figure 1. A single pipeline with a total length connected to a reservoir at the upstream end
and to a side discharge valve next to a dead end at the downstream end. The location of a potential leak is
described by the distance from the reservoir to the anomaly and the position of the side discharge valve
is fixed, always at the downstream end. The head measurements are also obtained at the downstream end of
the pipeline (denoted G/M since it corresponds to both thetransientgeneration and measurement point).
Figure 1. Pipeline configuration.
In order to obtain an accurate non-dimensional transformation ofthe head traces, no previous assumptions
are made in the numerical modeling with regard to the initial conditions of the system. The steady state
hydraulics of the pipeline are represented in Figure 2. In general, the flow through a leak and the flow through
the side discharge valve, before the generation of the transient event dependson the HGL along the pipeline.
In the same way, the resulting HGL depends on the outflows.
Leaks have been modeled as circular orifices with fixed diameters and a discharge coefficient of 0.6 that
dischargesto the atmosphere. Therefore, the flow through the leak can be expressed as a function of the head
at the location of the leak. In addition, the side discharge valve was modeled as an elementto discharge water
into the atmosphere with an energy loss coefficient of 0.05 when fully opened. Given that this side discharge
valve can be partially open at the steady state (to simulate the creation of a small transient event), reference
values for the valve are found using the maximum flow through the pipeline and a valve totally opened.
229 |
ADE | E-proceedings of the 38th IAHR World Congress
September 1-6, 2019, Panama City, Panam
Figure 2. Pipeline steady state modeling.
Finally, the outflow at the reservoir is calculated by adding the flow through the leak and the side discharge
valve. For the examplesshowna remaining base flow was discarded but if the system changes to a continuous
pipeline (instead of a dead end), the outflow at the reservoir would include this base flow. The steady state
modeling proposed forms a system of three equations and three unknowns. The three unknowns are the flow
at the reservoir, the head at the leak location and the head just upstream the side discharge valve. The three
equations are the head loss in the segment of pipe upstream the leak, the head loss in the downstream
segments of the pipe and the continuity equation. These three non-linear equations can be solved using any
numerical method. However, MATLAB embedded functions were used in this case which includes trust-region
and Levenberg-Marquardt algorithms(MathWorks 2019).
4 ANON-DIMENSIONALTRANSFORMATION
The purpose of the non-dimensional transformation of a head trace is to obtain a method to transform any
head trace that includes a leak response into a standardized form that allows its characterization. This
transformation is proposed in two parts: the transformation of the head trace and the non-dimensional
characterization of the location and the size of the leak. By applying this transformation, the head trace of two
different pipelines would look the same if the leaks in both have the same non-dimensional characteristics.
4.1 Headtransformation
Previous authors have normalized the head in a pipeline by establishing the ratio between the head at any
moment of time and a reference head (Wang et al. 2002). However, the normalization proposed in this paper
considers two different reference heads: the steady state head and the initial head rise after the closure of the
side discharge valve.
[1]
The proposed non-dimensionalization is shown in Equation 1. is the original head trace and
represents the non-dimensional head. First, the steady state head in the pipeline is subtracted to obtain
only the variation of head in the pipeline after the transient event. In addition, this variation of pressure is divided
by the initial head rise. By doing this, the final normalized head is always a proportion of this initial head rise.
4.2 Timetransformation
In order to normalize time, the period of a reservoir-pipeline-valve system was used. A complete cycle of
reflections after the water hammer event will happenin the seconds after the closure of the side discharge
valve. Therefore, the normalization of time is carried out by dividing the time in the original head trace by this
value.
[2]
Equation 2 presents the non-dimensionalization in time. Previous authors have normalized time by dividing
time only by ; however, since some pipe condition assessment techniques work with the first cycle of
reflections; itis useful to have the time normalized in cycles that correspond to the period of the pipeline.
These two equations above transform the head trace into a non-dimensional form and are valid for any
head trace after the generation of a transient event, no matter the anomaly that might be present or even if is
an intact pipeline trace. However, by only doing these transformations it is not possible to characterize the
presence a leak or any other anomaly. To accomplish these, two more non-dimensional quantities are proposed.
4.3 LeakLocation
The first and most important characteristic about a leak or any other anomaly is its location. The most
popular way of normalizing the location of an anomaly, that has been used previously by other authors is dividing
the real location of the leak by the total length of the pipeline. In this paper the same approach is used and is
shown in Equation 3.
[3]
Following this equation, the non-dimensional location will always be a number between 0 and 1
representingthe percentage of the distance from the reservoir where the leak is located. This non-dimensional
location is valid for any anomaly, however in this paper focus will be only on the presence of leaks.
230 3 |
ADE | E-proceedings of the 38th IAHR World Congress
September 1-6, 2019, Panama City, Panama
4.4 Leak Characteristics
In addition to the location of the leak, the size of the leak is also important in the characterization of this
anomaly. Since the leak isusuallymodelled as a circular orifice, the diameter of this orifice would represent the
size directly. However, in a real pipeline the leak might not a circular orifice. Therefore, the characterization of
the size of the leak is proposed using the flow through the leak.
[4]
Equation 4 presents the non-dimensional size expression for the leak. As it was shown in Figure 1, is
the flow through the leak and is the flow through the side discharge valve before its closure to create the
transient events. A non-dimensional characterization using flows was proposed by Wang et al. (2002) using a
generic reference flow, however,in this paper the reference flow is taken to be the side discharge valve flow.
For the purpose of this paper, the leak is always modelled as a circular orifice. Therefore, from the flow
through the leak is possible to find the diameter associated with this flow based on Equation 5.
[5]
4.5 Application
The main purpose of the proposed non-dimensional approach is to transform a head trace based only on
basic information of the system and on the head measurements. For example, to normalize the head, only the
steady state and the initial head rise are necessary and these two can be obtained easily from the head trace.
It is important to highlight that the side discharge valve closure does not need to be instantaneous for this
normalization to be valid. What matters is the value of the head rise after the closure of the valve. For real
pipelines it is enough to takean average of different pressure values obtained after the closure of the valve.
For the normalization of the time, the total length of the pipeline and its wave speed are required. In some
cases, these characteristics of the pipeline might be known. However, in some other cases, their certainty is
unclear. The wave speed can be determined if different measurement points are installed by calculating the
travel time between two measurement points. Alternatively, the period of the pipeline can also be
obtained directly from the head trace by calculating the time of the first prominent reflection from the reservoir,
which happens at seconds.
Another requirement to characterize a leak based on the previous equations is the flow through the side
discharge valve. Measuring this value during a transient test might be challenging depending on the available
equipment. However, if the closure of the valve is rapid(not necessarily instantaneous) this flow can be obtained
using the head rise in the pipeline and the hydraulic impedance following Equation 6.
[6]
This equation is obtained by applying the Method of Characteristics (MOC) to a situation like the one
described in this paper. A similar situation is described inBohorquez et al. (2019) where the generation of the
transient wave is at any point of the pipeline. For this case, the side discharge valve is always at the downstream
end of the pipeline. Following this equation, only with the head rise in the head trace and the hydraulic
impedance (defined as byWylie and Streeter (1993)) it is possible to obtain an estimate of the flow
through the valve itis closed.
In the following sections, numerical and experimental examples are shown to demonstrate that the non-
dimensional transformation proposed is valid and allows the characterization of head traces after the generation
of a transient event and with the presence of a leak.
5 NUMERICAL VALIDATION
A first validation was developed using a numerical model of transient flow using the MOC. The main
purpose of the validation was to compare head traces that are generated from different pipelines with different
characteristics after the application of the non-dimensional transformation. All the examples shown in this
section correspond to the system configuration described in Figure 1, however, the specific dimensions are
different.
For demonstration purposes, the selected examples correspond to the same non-dimensional location and
size of a leak (Equations 3 and 4) in order to show that the resultsfrom the non-dimensional transformation is
equivalent. The characteristics of the pipelines for Examples 1 and 2 are presented in Table 1. As it can be seen
in the table, the length, diameter and initial conditions of both pipelines are completely different. In addition,
231 |
ADE | E-proceedings of the 38th IAHR World Congress
September 1-6, 2019, Panama City, Panam
Example 1 has been modelled with an instantaneous closure of the side discharge valve while Example 2
includes a closure time of 0.05 s.
Table 1. Properties of numerical examples.
PROPERTY EXAMPLE 1 EXAMPLE 2
HEAD AT RESERVOIR HR(m) 66 38
LENGTH L (m) 356 1265
DIAMETER D(mm) 427.40 211.60
VALVE DIAMETER D (mm) 75 20
LEAKDIAMETER D (mm) 34 12.82
LEAK LOCATION L leak(m) 136.40 484.68
WAVE SPEED a (m/s) 1166 1178
PIPE WAVE PERIOD 4L/a (s) 1.22 4.29
VALVE CLOSURE TIME Cl T(s) Instantaneous 0.05
Table 2 presents a summary of the hydraulic results of the transient simulation in both examples. As can
be seen, the flow through the valve and the leak are completely different in both examples mainly due to the
difference in the size of the side discharge valve (shown in Table 1). The initial head rise is also different in both
examples. The only two characteristics that are the same, as mentioned previously, are the non-dimensional
leak location and size.
Table 2.Hydraulic and non-dimensional propertiesof numerical
examples.
PROPERTY EXAMPLE 1 EXAMPLE 2
FLOW THROUGH VALVE QGen(L/s) 35.53 3.83
LEAK FLOW QL(L/s) 19.60 0.21
INITIAL HEAD RISE i(m) 29.44 13.08
NON-DIMENSIONAL LOCATION L*(-) 0.383 0.383
NON-DIMENSIONAL SIZE Size*(-) 0.552 0.552
The original head traces obtained from Examples 1 and 2 are shown in Figure 3. Both head traces have
beenplotted onthe same figure to show that the results are completely different. These differences are evident
not only because the steady state head is different but are due to the differences in pipe lengths, and the
development of the transient reflections after the side discharge valve closure which are completely different.
The length of the pipeline in Example 1 is shorter, therefore the reflections from the reservoir arrive to the
measurement pointisfaster and a complete cycle is achieved at an earlier time. This is also evident in Table 1
where the period of each pipeline is reported.
Figure 3.Head traces for Example 1and 2.
232 |
ADE | E-proceedings of the 38th IAHR World Congress
September 1-6, 2019, Panama City, Panama
With these results, the non-dimensional transformation proposed earlier in this paper was applied to
Example 1 and 2. Figure 4 presents the non-dimensional head traces for both examples. As it can be seen in
this figure, once the head and the timeare normalized, the two traces become equivalent, especially during the
first two periods after the closure of the side discharge valve. When the first seconds are analyzed
(corresponding to 1.1 approximately in Figure 4 given that the side discharge valve is not closed immediately
after the beginning of the simulation),itis possible to note that the normalization of the head allows for an easy
and quick comparison between the two traces and that the initial reflection of the leak looks almost thesame in
both examples. The differences come from the fact that in Example 1 the initial wave is completely sharp, while
in Example 2 the side discharge valve has a finite closure time. In the same way, the differences in the rest of
the trace come from the same fact and although these differences are visible, in general terms both traces are
quitesimilar.
Figure 4. Non-dimensional head traces for Example 1 and 2.
A third example has been developed to show the capability of the non-dimensional head trace to represent
different dimensional examples. Table 3 presents the characteristics and hydraulic results from a third example
that has the same non-dimensional leak location and size than Example 1 and 2. Once again,it can be seen
that the initial conditions and characteristics of the pipeline are different from the other examples. Given that
this pipeline is the shortest of the three examples, its period is only 0.31 s. The purpose of this example is to
show that any of the two non-dimensional head traces shown in Figure 4 can be transformed into a dimensional
head trace that would be very similar to the head trace obtained from Example 3.
Table 3. Properties of Example 3.
PROPERTY EXAMPLE 3
HEAD AT RESERVOIR HR(m) 55
LENGTH L (m) 89
DIAMETER D (mm) 573.40
VALVE DIAMETER D (mm) 80
LEAK DIAMETER D (mm) 45.90
LEAK LOCATION L leak(m) 34.10
WAVE SPEED a (m/s) 1158
PIPE WAVE PERIOD 4L/a (s) 0.31
VALVE CLOSURE TIME Cl T(s) Instantaneous
FLOW THROUGH VALVE QGen(L/s) 59.07
LEAK FLOW QL(L/s) 32.61
INITIAL HEAD RISE i(m) 27.00
NON-DIMENSIONAL LOCATION L*(-) 0.383
NON-DIMENSIONAL SIZE Size* (-) 0.552
The non-dimensional head trace from Example 1 shown in Figure 4 has been transformed into a
dimensional head trace by using Equations 3 and 4. The non-dimensional time of Example 1 was multiplied by
the period of Example 3and the non-dimensional head was multiplied by 27.00m and the steady state head of
Example 3 was added (these values are both shown in Table 3). Once the dimensional head trace was obtained,
it was compared with the head trace obtained from the simulation of the transient flow model and the results
are presented in Figure 5.
233 |
ADE | E-proceedings of the 38th IAHR World Congress
September 1-6, 2019, Panama City, Panam
Figure 5. Head trace for Example 3 (Generated with MOC and
generated from a different non-dimensional trace).
By inspecting Figure 5 is possible to see that the head trace obtained from the non-dimensional trace of a
different system matches the head trace obtained from the numerical model, especially at the beginning of it.
The first reflection of the leak is almost identical. There are some differences in the trace after two complete
cycles of reflections mainly due to differences in the frictional energy dissipation because the equations
proposed to transform the traces do not include friction. However, if only the first two cycles are used to identify
the anomaly (in this case a leak), the transformation is successful.
The numerical validation presented demonstrate that the non-dimensional transformation is useful to obtain
standardized head traces that are independent of the characteristics of the pipeline. By accomplishing this,
techniques to locate anomalies that are general and do not require specific information about the analyzed
system can be proposed and applied. The purpose of this paper, however, is to show that the non-dimensional
transformation is valid and general.
6 LABORATORY VALIDATION
A set of laboratory tests have been used to validate the non-dimensional transformation proposed in a
system that is different from the numerical examples shown above. A total of 4 experimental tests are shown
where the configuration of the test corresponds to the one described in Figure 1. All tests were developed in the
Robin hydraulics laboratory of The University of Adelaide. The first two tests were developed in 2005 as part of
a previous research (Lee 2005) and the last two tests were conducted in the same pipeline for the purpose of
this validation.
6.1 Laboratory Configuration
A schematic view of the pipeline is presented in Figure 6. The apparatus includes a straight 37.39 meters
long copper pipeline with an internal diameter of 22.1 millimeters and 1.6 millimeters wall thickness. The
difference in elevation between the two ends is 2 meters. The pipeline is connected to electronically regulated
pressure tanks with in-line valves for flow control to give the apparatus the ability to simulate different boundary
conditions. Depending on the test, one of these in-line valve was completely shut to simulate a reservoir-pipeline
valve system.
To simulate the leak, side-discharge orifices were connected at certain points along the pipeline. These
possible locations are shown in Figure 6 as points 1, 2 and 3. The transient events were generated at points
marked as G1 and G2 in Figure 6 given the installation of a side-discharge solenoid valve which can be closed
in approximately 4 milliseconds.
Figure 6. Laboratory setup.
6.2 Tests
Using the pipeline described above, four tests were used to validate the non-dimensional transformation
of head traces. Figure 7 presents the location and the size of the leak for the four tests. As it can be seen, the
leak size was varied between 1.0 and 3.0 mm.In addition, the location of the solenoid valve used to simulate a
234 |
ADE | E-proceedings of the 38th IAHR World Congress
September 1-6, 2019, Panama City, Panama
side discharge valve was changed between the two ends of the pipeline to create different examples. The head
traces for test 1 and 2 were obtained from previous research conductedat the University of Adelaide (Lee 2005).
Tests 3 and 4 were conducted in 2018 in the same pipeline with different flows through the main pipeline to
illustrate the usefulness of the non-dimensional transformation.
(a) (b)
(c) (d)
Figure 7. Configurationsof laboratory tests (a) Test 1, (b) Test 2, (c) Test 3 and (d) Test 4.
The value for the wave speed was obtained in two ways. In tests 1 and 2, the wave speed was obtained
from the results reported by Lee (2005) as 1328 m/s. In test 3 and 4 it was calculated from the head
measurements at two points of the pipeline from the time difference of the arrival of the first wave front. From
thesecalculations the obtained wave speed was validated to be1328 m/s.In addition, from the measurements
the closure time of the solenoid valve was estimated to be 4 milliseconds. Table 4 summarizes the main
characteristics of the tests including the non-dimensional location and size of the leaks in those tests.
Table 4.Characteristics of Laboratory Tests.
CHARACTERISTICS TEST 1 TEST 2 TEST 3 TEST 4
HEAD AT RESERVOIR HR(m) 40 37 27.84 28.20
LEAK DIAMETER D (mm) 1.50 1.0 2.0 3.0
INITIAL HEAD RISE i(m) 17.70 17.92 12.92 12.96
FLOW THROUGH VALVE QGen(L/s) 0.0503 0.0510 0.0367 0.0369
LEAK FLOW QL(L/s) 0.0297 0.0127 0.0441 0.0998
NON-DIMENSIONAL LOCATION L*(-) 0.178 0.752 0.297 0.297
NON-DIMENSIONAL SIZE Size* (-) 0.590 0.249 1.199 2.707
Table 1 shows that even though the tests were conducted in the same pipeline, the initial conditions and
the non-dimensional characteristics of every test were different. Is important to note that the flow through the
solenoid valve and the flow through the leak were calculated from the head traces using Equations 5 and 6.
6.3 Results
Results from the four tests are shown in Figure 8 and 9. Figure 8 presents the dimensional head traces for
the first 0.14 secondsof the test which corresponds almost to a complete cycle of reflections seconds).
In this figure,it can be seen that the response of the leak in Tests 1 and 2 is more subtle, mainly because of the
size of the circular orifice that represents the leak. In both cases (Figure 8 (a) and (b)), the drop in pressure
characteristic of the reflection of the leak is almost imperceptible. For tests 3 and 4 the leak is more prominent
given that the size is larger. Even in test 4 some perturbance in the head is visible along the whole test mainly
because a leak size diameter of 3.0 millimeters is disruptive in a pipeline of 22.1millimeters of inner diameter.
235 |
ADE | E-proceedings of the 38th IAHR World Congress
September 1-6, 2019, Panama City, Panam
(a) (b)
(c) (d)
Figure 8. Head traces laboratory tests (a) Test 1, (b) Test 2, (c) Test 3 and(d) Test 4.
In order to validate the non-dimensional transformation, non-dimensional numerical head traces were
generated and compared with the non-dimensional version of the traces from the laboratory tests. In a similar
way as in the case of the numerical validation, the non-dimensional location and sizes of the leak were preserved
for demonstration purposes. This comparison is presented in Figure 9 for all the tests.
In general, the head traces obtained from the laboratory matched successfully the numerically generated
traces once the non-dimensional transformation is applied. For all tests, the moment in time at which the
pressure drop due to the presence of the leak happens isthe same time (as evidence of the location of the leak)
and the non-dimensional magnitude of the drop is also similar for both cases (as evidence of the size of the
leak). Even in test 4, where the head obtained from the laboratory displays some perturbance, the non-
dimensional transformation allows the characterization of the occurrence of this leak.
(a) (b)
(c) (d)
Figure 9. Non-dimensional traces. Laboratory and numerical results. (a) Test 1, (b) Test 2, (c) Test 3 and
(d)Test 4.
236 |
ADE | CHAPTER 1
Regional and Local Geology and the Aims of this Study
1.1 Introduction
The world’s largest Zn-Pb-Ag deposit (Figure 1.1), Broken Hill, in far western New
South Wales, was discovered in September 1883 by a German noble man who called himself
Charles Rasp (Sainisch-Plimer 1999). The Broken Hill orebody is in the Palaeoproterozoic
Curnamona Province, formed at 1686 ± 5 Ma, underwent multiphase intense deformation and
high grade metamorphism at 1600 Ma (Olarian Orogeny) and less intense metamorphism in
the 500 Ma (Delamerian Orogeny). It is an 8 km long linear mass that has the shape of a
boomerang that plunges both NE and SW. It crops out at the centre where some 80 Mt of
sulphide rocks have been lost by weathering and erosion (Plimer 1984). The orebody has
produced over 250 million tonnes of high-grade ore originally containing 28 Mt Pb, 24 Mt Zn
and 1 billion oz1 Ag with estimated revenue in today’s Australian dollars of $ 300 billion. The
wealth generated from Broken Hill led to the industrialisation of Australia and the formation
of the world’s two largest mining companies, BHP Billiton and Rio (Blainey 1968). Satellite
images of the Broken Hill Mine are provided in supplementary files to this thesis.
The Broken Hill deposit is associated with hundreds of small Broken Hill-type
deposits in the Willyama Supergroup (Stevens 2003). The main Broken Hill deposit and
hundreds of Broken Hill-type deposits have the same lead isotope signature thereby
suggesting similar formation processes (Gulson et al. 1985; Stevens 2003) whereas the small
Pinnacles Zn-Pb-Ag deposit has a Pb isotope signature suggesting that it is some 10 Ma older
than the main Broken Hill deposit (Parr, Stevens & Carr 2003) yet formed by the same
process. The Broken Hill ore deposit is in a world-class metallogenic province wherein there
more than 4,000 pits and shafts have been sunk for a diversity of commodities such as Zn, Pb,
Ag, Cu, Au, W, Sn, Be, U and non-metallic minerals. The world’s largest zinc-lead deposit is
in a province where most rocks are enriched in Zn and Pb, sulphide deposits of a number of
types are common and soils derived from sulphide-bearing rocks and silicate rocks produce a
metal-enriched regolith.
1 Ounces |
ADE | 2 Chapter 1-Regional and Local Geology
NOTE:
This figure is included on page 2 of the print copy of
the thesis held in the University of Adelaide Library.
Figure 1.1: Contained grade and tonnage of metals for stratiform Zn-Pb deposits.
The Broken Hill deposit is the largest Zn-Pb known deposit in the world
(redrawn from Large et al. 2005).
1.2 Curnamona Province
The Curnamona Province extends across eastern South Australia and western New
South Wales (Figure 1.2). It is ovoid in shape, has an area of about 50,000 km2 and is
divided into eight domains (Conor & Preiss 2008). It comprises Palaeoproterozoic
metasediments, metavolcanics and chemical metasediments (Willyama Supergroup)
intruded by early Mesoproterozoic granites and, in places, covered by early
Mesoproterozoic volcanic rocks (Robertson et al. 1998). Neoproterozoic sequences
unconformably overlie the older rocks of the Curnamona Province. There is a temporal
stratigraphic correlation (~1720–1640 Ma) between the Broken Hill and the Olary
Domains and many geochronological studies have correlated stratigraphic units between
the two Domains (e.g. Clarke, Burg & Wilson 1986; Willis et al. 1983). |
ADE | Chapter 1-Regional and Local Geology 3
NOTE:
This figure is included on page 3 of the print copy of
the thesis held in the University of Adelaide Library.
Figure 1.2: The eight Domains of the Curnamona Province (from Conor et al. 2006).
PIRSA have compiled 1:25,000 sheets of the Olary Domain collated from a 10-year
integrated 1:10,000 outcrop lithological and structural mapping program by honours
students from The University of New England, The University of Newcastle and The
University of Melbourne. There has been limited mapping of the Moolawatana Domain by
students and much of the mapping has been undertaken by Teale (unpublished). Because of
the lack of detailed geochronological information and absence of some important
stratigraphic successions of the Broken Hill Group within the Olary Domain (e.g. Parnell
Formation; Freyers Metasediments; Hores Gneiss; Figure 1.3; Page et al. 2005), the
stratigraphic correlations are somewhat loose. It is the unit that contains Broken Hill-type
mineralisation that is absent from the Olary Domain and possibly the Moolawatana
Domain.
With U-Pb geochronological dating of detrital zircon, the age constraints of
depositional sediments and important magmatic and metamorphic events are limited to
predominantly felsic rocks and the timing of mafic rocks may not be determined precisely
as some may be intrusive (Yamashita, Creaser & Villeneuve 2000). This is the case in the |
ADE | Chapter 1-Regional and Local Geology 5
Some different stratigraphic successions in the Broken Hill Domain have
overlapping detrital zircon ages, despite the fact that they are not juxtaposed. This may
indicate unconformities. For example, quartz-albite rocks sequences of the Himalaya
Formation in the Thackaringa Group have overlapping detrital zircon ages with major
granitic gneiss sills adjacent to the Broken Hill ore deposit yet the two units are spatially
separate (Raetz, Krabbendam & Donaghy 2002). Inherited refractory zircon cores in
magmatic zircon grains result from partial to complete melting of older sediments or
magmatic rocks. The dominant population of inherited zircon ages derive from weathering
of older source rocks, magmatism or xenocrysts hence the maximum age of deposition
cannot be precisely constrained because there is no certainty that the samples analysed
contain the youngest zircon present in a population (Page & Laing 1992). Although
unconformities, disconformities and layer-parallel thrusts have not been recognised at
Broken Hill, this does not mean they do not exist. Notwithstanding these limitations, for
the purposes of this study, the Broken Hill Domain geochronology used is that of Raetz,
Krabbendam & Donaghy (2002) is used.
The Geological Survey of New South Wales constructed a stratigraphy and later
validated this with geochronology (Figure 1.3). This stratigraphy is constantly being
revised (albeit slightly), has a predictive basis and can be used in the field. Its strengths are
that the mapping was fact outcrop mapping done by an integrated team who used the same
lithological nomenclature. Because of lack of exposed basement and erosion of the top of
the Willyama Supergroup in the Broken Hill Domain, its overall thickness is not known
precisely and it is estimated between 6-7 km (Stevens et al. 1983; Willis et al. 1983).
SHRIMP geochronological studies of detrital zircons (Page et al. 2005; Page & Laing
1992; Page, Stevens & Gibson 2005; Raetz, Krabbendam & Donaghy 2002) and
stratigraphy (Willis et al. 1983) show deposition of the Willyama Supergroup took place
from ≤ 1710 Ma at the base to ≤ 1642 ± 5 Ma at the top (Nutman & Gibson 1998; Page,
Stevens & Gibson 2005; Stevens 2000; Stevens et al. 1988).
The weakness of the NSW Geological Survey stratigraphic mapping is that there
was no structural mapping hence deformational interpretation is based on lithology and not
measured structural fabrics. Furthermore, the Parnell Formation was mapped as a
stratigraphic unit rather than Allendale Metasediments intruded by mafic sills at a |
ADE | 6 Chapter 1-Regional and Local Geology
buoyancy level. However, the mafic rocks which define the Parnell Formation are intrusive
(James, Pearce & Oliver 1987; Phillips, Archibald & Wall 1985; Stevens 2009). Stevens
(1998) argued that it is not clear whether the amphibolite rocks of the Parnell Formation
occurred during deposition of the Willyama Supergroup or younger deposition or whether
the amphibolite rocks are related to sub-volcanic events. More recent geochronology
shows that the amphibolites of the Parnell Formation (Figure 1.3) were intruded as high
Fe- and Ti-tholeiitic basalts (Raveggi et al. 2007) into wet sediments at 1685 ± 5 Ma
thereby establishing geothermal systems (Plimer 2006b; Stevens 2003).
1.3 The Willyama Supergroup
Over a fifteen-year period, the Geological Survey of New South Wales mapped the
Broken Hill Domain at 1:12,500 scale maps and published geological maps at 1:25,000
scale. Supporting geophysical, regolith, metallogenic and summary maps were published at
1:25,000, 1:50,000, 1: 100,000, 1:250,000 and 1:1,000,000 scales. The Willyama
Supergroup in the Broken Hill Domain is dominated by rocks interpreted as
metamorphosed highly deformed clastic metasediments (Andrews 1922; Gustafson,
Burrell & Garretty 1950; Willis et al. 1983) that formed in an intra-continental rift
succession (Willis et al. 1983). The metasediments show features interpreted as
sedimentary structures such as bedding, graded bedding, cross bedding, ripples and
pull-apart structures (Laing, Marjoribanks & Rutland 1978; Slack et al. 1993; Willis et al.
1983). The metasediments are intruded by pre-, syn- to post-deformational felsic and mafic
magmatic rocks and, where magmatism is bimodal at 1686 ± 5 Ma, there are occurrences
of Broken Hill-type Zn-Pb-Ag mineralisation and associated pre-metamorphic alteration
(Conor & Preiss 2008; Parr & Plimer 1993; Phillips, Archibald & Wall 1985; Plimer 1975,
1986).
The Broken Hill orebody is hosted by the Broken Hill Group, a unit that varies in
thickness from 300 to 2000 m, but mostly between 1000 m and 1500 m. The depositional
age of the Broken Hill Group is estimated between the ages of the Plumbago Formation
(1693 ± 3 Ma) in the Olary Domain and the Hores Gneiss (1685 ± 3 Ma) in the Broken
Hill Group (Page et al. 2005). The main conclusions from studies of the Broken Hill
Domain (Conor & Preiss 2008; Laing 1996a; Phillips, Archibald & Wall 1985; Plimer
1985; Slack & Stevens 1994; Stevens 1986; Stevens et al. 1988; Willis et al. 1983) were
that the Broken Hill orebody was deposited in a deep-water intra-cratonic rift. |
ADE | Chapter 1-Regional and Local Geology 7
Since that time, Stevens (2003), Feldman (2004), Plimer (2006b) and Damm (2008)
have reinterpreted the area as a shallow water intra-continental rift setting on the basis of
detailed sequence stratigraphy of the metasediments. Plimer (2006b) suggested that the
depositional setting was a shallow fresh water intra-cratonic continental rift lake. Outside
the precinct of the orebody, the uppermost unit of the Broken Hill Group (Hores Gneiss) is
dominated by felsic gneiss whereas the Hores Gneiss enclosing the Broken Hill orebody is
thicker, dominated by clastic metasediments and has very minor units of felsic gneiss. This
may support the view that the Broken Hill orebody formed in a depression or rift (Plimer
2006b). Although major transgressive retrograde shear zones were mapped, it is surprising
that no thrusts were recognised because it is expected that nappe (F ; Laing, Marjoribanks
1
& Rutland 1978) and isoclinal (F ; Laing, Marjoribanks & Rutland 1978) folding would
2
produce thrusts.
1.4 The Broken Hill ore deposit
The Broken Hill ore deposit is hosted by metasediments that have undergone
variable degrees pre-metamorphic hydrothermal alteration to quartz-, garnet- and gahnite2-
rich assemblages (Groves et al. 2008; Haydon & McConachy 1987; Heimann et al. 2009;
Plimer 1979). The initial fluid conduit for the Broken Hill ore deposit is in the southern
part of the field between The Zinc Corporation and New Broken Hill Consolidated Mines
(Groves et al. 2008; Plimer 1979). The ore deposit (Figure 1.4) comprises the following
Lodes and Lenses:
C Lode: a Zinc Lode and a transgressive footwall alteration zone (Groves et al.
2008; Plimer 1979),
B Lode: a basal Zinc Lode (Groves et al. 2008; Hodgson 1974; Johnson &
Klingner 1975; Plimer 1979; Segnit 1961; Stilwell 1959),
A Lode: two manganoan Zinc Lodes (Groves et al. 2008; Johnson & Klingner
1975; Plimer 1979),
1 Lens: a Zinc orebody (Groves et al. 2008; Johnson & Klingner 1975),
2 Lens: a Lead orebody (Groves et al. 2008; Johnson & Klingner 1975; Mackenzie
1968; Plimer 1979), and
2 (Zn, Fe) Al O
2 4 |
ADE | 8 Chapter 1-Regional and Local Geology
3 Lens: a Lead orebody and an upper most mass (Groves et al. 2008; Johnson &
Klingner 1975; Plimer 1979).
In Broken Hill, the Silver City, was established on the secondary silver ore derived
from 2 and 3 Lens and it was only late in the history that the Zinc Lodes were exploited.
Stratigraphically equivalent to the Zinc Lodes is the Western Mineralisation, a poorly
defined low- to medium-grade sulphide mass discovered in 1913 by BHP3 (Blampain &
Plimer 2006; Gentle 1968; Plimer 2006b). In the 1950s, Broken Hill South cut a haulage
drive from Browne Shaft to No 7 Shaft through the Western Mineralisation for sampling,
drilling and access and, in the 1970s, MM&M4 Ltd. extracted a 1,500 tonne Western
Mineralisation ore parcel for trial milling. Stratigraphically equivalent to the Western
Mineralisation on the eastern limb of the Broken Hill Antiform is a quartz-, garnet-,
gahnite and sulphide-bearing zone (Eastern Mineralisation) that has yet to be explored
(Plimer 2006b). It is the Western Mineralisation that is the subject of this study, it is open
at depth, it has high-grade zones, possibly at zones of dilation (Plimer 2006b) to the south
and to the north and mining will commence in 2011 at the CBH Resources Ltd Rasp Mine.
B Lode, A Lode, 2 Lens and 3 Lens comprise the greatest tonnage of sulphide rocks at
Broken Hill (Table 1.1).
Table 1.1: Size and grades of the Broken Hill orebodies and the Western Mineralisation
(Plimer 2006b; Stevens 2003).
Orebody Tonnage Lead grade Zinc grade Silver grade
3 Lens 79 Mt 14 % Pb 14 % Zn 250 g/t Ag
2 Lens 85 Mt 14 % Pb 11 % Zn 100 g/t Ag
1 Lens 10 Mt 8 % Pb 20 % Zn 50 g/t Ag
A Lode 53 Mt 4 % Pb 10 % Zn 40 g/t Ag
B Lode 46 Mt 5 % Pb 17 % Zn 40 g/t Ag
C Lode 11 Mt 3 % Pb 5 % Zn 20 g/t Ag
The Western
17 Mt 2.2 % Pb 3.2% Zn 28 g/t Ag
Mineralisation
3 Broken Hill Proprietary Company
4 Minerals, Mining and Metallurgy Limited |
ADE | Chapter 1-Regional and Local Geology 9
NOTE:
This figure is included on page 9 of the print copy of
the thesis held in the University of Adelaide Library.
Figure 1.4: Open pits on CML75 from which carbonate, sulphate and sulphide ores were
extracted from 1883-2001 (from CBH Resources Ltd).
The BHP open pit is adjacent to Delprat Shaft and the Block 14 Pit is immediately
SW of the Blackwood Pit. During the 1980s, secondary minerals were mined from Kintore,
Block 14 and Blackwood Pits and primary sulphide ores were mined from No 7 Shaft,
Blackwood Pit and Browne Shaft. Tribute mining in the 1990s of secondary and primary
minerals in Block 14 Pit was undertaken by Mr. Craig Williams of Pinnacles Mines Pty
Ltd. This study concentrates on the upper part of the Western Mineralisation on CML7.
1.4.1 Mine sequence stratigraphy
A different stratigraphy for the metasediments enclosing the Broken Hill orebody
has been compiled by the mining companies because the regional geological mapping was
not undertaken in the mines area, the scale of regional mapping was different from the
scale of mapping in the mines area and there are subtle facies relationships associated with
the Broken Hill orebodies, although hosted by recognisable stratigraphic units (Tables 1.2
and 1.3). Geological cross-sections of the Broken Hill orebodies have been provided in
supplementary files to this thesis. The Broken Hill orebodies occur in Unit 4.7 and
5 Consolidated Mining Lease Number 7 or the Broken Hill Rasp Project |
ADE | 12 Chapter 1-Regional and Local Geology
segment of the Broken Hill orebody (Feldman 2004). Haydon and McConachy (1987) used
conventional mining company diamond drill core logging of lithology, geotechnical and
engineering aspects combined with down-hole gamma ray logging. During high grade
metamorphism, there is a decrease in porosity and permeability; there is massive
dewatering which removed K and U from Broken Hill rocks (Ahmad & Wilson 1982), two
generations of K- and U-rich pegmatites formed during the Olarian Orogeny (Laing,
Marjoribanks & Rutland 1978) and retrogression has resulted in the removal of K (Phillips
1980). Despite these problems, Haydon and McConachy (1987) were able to validate the
stratigraphy obtained from conventional logging and they suggested that the Broken Hill
orebody formed by the replacement of shallow water deltaic sediments.
In the Western Mineralisation, metasediments show sharply truncated graded
bedded stratigraphy, characterised by coarse grained sillimanite-almandine11 at the top and
fine grained quartz-feldspar at the base. Kitchen (2001) showed from only one Western
Mineralisation diamond drill core that the thickness of metapsammitic beds increased up
stratigraphy beneath the Western Mineralisation. Feldman (2004) tried a more
metapelite
conventional sequence stratigraphic approach and measured the ratio
metapsammite
metre-by-metre in 20 Western Mineralisation diamond drill cores. Feldman’s work was
limited by the number of available drill cores at the time; he attempted to establish
structure contours for one of the stratigraphic marker horizons (Unit 4.6 metapelite) and
attempted to deduce the environment of deposition for the sequences beneath the Broken
Hill orebody.
Despite limited data, Feldman (2004) concluded that there were three upward-
coarsening sediment cycles in the Broken Hill Group and that metapsammite horizon also
became thicker towards the top of each cycle. This has been validated by additional core
logging by Plimer (2006b) and Damm (2008). Massive metapelite of Unit 4.6 is
characterised by high a Al content (crenulated sillimanite, biotite, almandine), no quartz
and no feldspar. Unit 4.6 metapelite is used as sedimentary marker horizon (Blampain &
Plimer 2006; Haydon & McConachy 1987; Plimer 2006b). It contains a remarkably
stratigraphically continuous thin marker BIF with quartz, almandine, magnetite, gahnite,
hyalophane and sulphides. The stratigraphic bottom of Unit 4.6 (and 4.8) is very sharp
boundary and it is different from other lithological boundaries in the Broken Hill Group
11 Fe Al (SiO )
3 2 4 3 |
ADE | Chapter 1-Regional and Local Geology 13
which tend to be gradational or cyclical. This suggests a sudden change in depositional
environment which may be coincident with sudden deepening, formation of depositional
basin, intrusion of amphibolite and the establishment of geothermal system (Plimer
2006b).
1.4.2 Rock types of the Broken Hill ore deposit
1.4.2.1 Metasediments
Metasediments of the Broken Hill deposit comprise metamorphosed clastic
sedimentary rocks with minor chemical sediments such as chert, oxide-facies, BIF, pelite,
psammite, psammopelite, much interlamination, bedding, graded bedding and other
sedimentary structures. Major metasediments of the Broken Hill Group contain HFSE12,
Y13 and REEs14 that originated from the erosion of anorogenic granite and rhyolitic to
rhyodacitic rocks with A type chemistry (Slack & Stevens 1994). Local stratigraphic
sequences of quartz-feldspar ± biotite ± garnet gneisses in the Broken Hill Group (e.g.
Potosi-type gneiss) are assumed as source of the metasediment (Slack & Stevens 1994).
In Broken Hill region, Hores Gneiss dominated by felsic gneiss (Stevens & Barron
2002) whereas, in the mines area, the Hores Gneiss dominated by upward coarsening
cycles of sediments and very minor felsic gneiss. This suggests that the mines area was a
topographic low, graben or rift into which clastic sediment was deposited whereas the
Hores Gneiss in the Broken Hill region was an area with little clastic deposition. The main
constituents of metapelite, metapsammites and metapsammopelites in the Broken Hill
comprises sillimanite, quartz, biotite, ± garnet, ± K-feldspar, ± plagioclase and ± cordierite.
These rocks are more chloritised down-hole (i.e. up stratigraphy). Metapelites contain
much more sillimanite and mica in comparison with metapsammopelite and
metapsammites.
1.4.2.2 Pegmatite
The main constituents of pegmatites comprise quartz, ± feldspar (plagioclase and
K-feldspar), ± garnet, ± biotite, ± chlorite, ± plumbian orthoclase. Pegmatites may be
foliated, unfoliated, boudinaged and brecciated.
12 High Field Strength Elements
13 Yttrium
14 Rare Earth Elements |
ADE | 14 Chapter 1-Regional and Local Geology
1.4.2.3 Lode horizon rocks
Lode horizon rocks typically comprise the following rock types:
1. Blue quartz lode comprising blue quartz, ± gahnite, ± minor garnet,
2. Quartzite lode comprising quartz , ±gahnite, ± garnet, ± feldspar, ± biotite and ±
tourmaline, and
3. Garnet quartzite, garnetite and garnet envelope (Spry & Wonder 1989; Wonder,
Spry & Windom 1988, p.228), medium-coarse grained rocks contains Mn-rich
almandine, ± gahnite, ± biotite and ± feldspar. The altered garnet-bearing rocks
contain a variety of minerals such as blue quartz, hedenbergite15, gahnite and
sulphide minerals.
A package of metasediment, garnet quartzite, garnetite, blue quartz-gahnite lode
and pegmatite are spatially associated with all sulphide orebodies of the Broken Hill
deposit (Burton 1998; Johnson & Klingner 1975).
1.4.2.4 Amphibolite
Amphibolites were high Fe- and Ti-tholeiitic basalts typical of triple point magmas
in the mid ocean ridge and they show in situ differentiation (Brick 2005; Brown et al.
1983; James, Pearce & Oliver 1987) from feldspathic tops to more mafic base in accord
with younging directions in the metasediments. Amphibolites have also undergone pre-
metamorphic hydrothermal alteration (Phillips, Archibald & Wall 1985).
1.4.2.5 Potosi Gneiss
Potosi Gneiss is a local name for foliated garnet-feldspar-biotite-quartz bearing
gneiss (Raetz, Krabbendam & Donaghy 2002). It occurs in both Unit 4.4 and 4.7. Plimer
(2006b) argued that Potosi Gneiss proximal to the Broken Hill mines may be an altered
rock in comparison with other lithological units of Hores Gneiss in Unit 4.7. In Broken
Hill Mine, there is trend from the garnet-bearing feldspathic psammite and the garnet-
plagioclase gneiss into blue quartz -garnet ± biotite ± pyrrhotite ± chalcopyrite rock. The
Potosi Gneiss appears in the lowermost part of the Parnell Gneiss which overlies the
Allendale Metasediments and is overlain by the Freyers Metasediments (Conor & Preiss
2008).
15 CaFeSi 0
2 6 |
ADE | Chapter 1-Regional and Local Geology 17
F folds are coeval lower grade metamorphism because of occurrence of
3
muscovite-defined schistosity suggesting isobaric cooling (Page & Laing 1992; Stevens
1986). Haydon and McConachy (1987) and Laing (1996b) suggest F remobilised
3
sulphides whereas Plimer (1984) suggests that F is the last phase whereby there is major
3
sulphide remobilisation were remobilised in even later events.
Table 1.6: Summary of the major deformation (D), metamorphism (M), schistosity (S) and
P-T conditions during D of the Olarian Orogeny.
3
Event Event Deformation Structure and Mineral P-T
Name label features foliation assemblage conditions
Development
retrograde Shear
zones
(Laing, Marjoribanks
& Rutland 1978).
Development of Development of Greenschist
minor F folds quartz, sericite facies
3 Development of the
with open to and biotite T=550-600°C
Globe Vauxhall,
monoclinal folds assemblages a n d
Retrograde British,
close to retrograde within shear P=5-5.5 kbar
De Bavay Shear zones
shear zones zones ( P h i l l i p s
(White et al. 1995)
(White et al. 1995). (Morland & 1980)
Webster 1998).
Cross cutting the
Main Line of Lode
(White et al. 1995)
and the Main Shear
(Rothery 2001)
At least two types of dolerite dykes dated at 830 Ma (dyke swarms and gabbroic
intrusions) are known in the Broken Hill Domain (Stevens et al. 1988). These are the same
age as the dyke swarms in South Australia (Gairdner Dyke Swarm) and the same age as the
initiation of the breakup of Rodinia (Li, Zhang & Powell 1995; Zhao & McCulloch 1993).
This suggests the Broken Hill and Olary Domains underwent extension at this time. In the
Broken Hill orebody, drill holes have intersected epidotised tholeiitic dolerite dykes in the
central part of deposit, including the Western Mineralisation, and Unit 4.7 at depth
)aM
0051-0061~(
ynegorO
nairalO
)5991
.la
te
etihW(
)aM
01±
6551(
D
3
)3002
tgreH
&
daehdooW
,illenoT( |
ADE | 18 Chapter 1-Regional and Local Geology
(Plimer 2006a). The sample cores of the Western Mineralisation within Unit 4.7 shows N-
and NW- trending epidotised tholeiitic dolerite dykes. Rare sulphide minerals were present
within the dolerite samples that transgress the Western Mineralisation. Dolerite samples
were not considered in detail for this study. Dolerite dykes intruded massive sulphide rocks
of 2 and 3 Lens resulting in fragmentation and garnetisation (Plimer 1984).
1.5.2 Delamerian Orogeny
The Delamerian Orogeny (Table 1.7) consists of two folding events (D and D )
5 6
coeval low grade metamorphism (Corbett & Phillips 1981). In the high grade metamorphic
grade rocks of the Broken Hill Domain, the high grade shear zones may have been
rejuvenated, new shear zones may have been initiated and there may have been a little
hydration and resultant retrograde metamorphism of silicate rocks. Shear zones that cut the
orebody (e.g. British Shear Zone) displace the orebody and remobilise Ag and Pb towards
the shears (Plimer 1984). Sulphide rocks with tholeiitic dolerite clasts shows that the
orebody must have moved during the Delamerian Orogeny (Plimer 2006b).
Table 1.7: Summary of the major deformation (D), metamorphism (M), schistosity (S) and
P-T conditions during D and D of the Delamerian Orogeny.
5 6
Event Event Structure Mineral
Deformation features P-T conditions
Name label and foliation assemblage
Reactivation of
retrograde shear zone
associated with post D
4
Development of (Phillips 1980) and cross
Weak siderite, Greenschist
cutting pre-existing shear
foliation muscovite, facies
structure.
developed chlorite, metamorphism
Brittle fault system
throughout sericite and T=350°C
developed within the
the deposit q u a r t z a long and
mineralised zone
(White et al. faults, shears, P=2 kbar
(Webster 1996).
1995). vugs and joints (Phillips 1980)
Sulphide minerals were (Morland &
remobilised through the Webster 1998).
reactivation of retrograde
shear zones as veins
(White et al. 1995).
)aM
584-025~(
ynegorO
nairemaleD
)3002
hcivoraB
&
drofrehtuR
,dnaH(
D
&
D
6
5
)1891
spillihP
&
ttebroC( |
ADE | Chapter 1-Regional and Local Geology 19
1.6 Characteristics of the Broken Hill orebodies
The Broken Hill orebodies comprise four Zinc ores (Zn > Pb) and two Lead
ores (Pb > Zn) as well as some non-outcropping mineralised zones (e.g. the Western
Mineralisation and the Centenary Mineralisation). These are closely correlated with each
other over a strike length of more than 8.5 km along the Main Line of Lode and some
minor discontinuous orebodies occur over a strike length of 25 km. On scales of 1:2000
and smaller, the Broken Hill ore deposits are considered as stratiform or stratabound that
underwent major and minor multiphase deformation (Plimer 2006b). Lawrence (1968),
Walters (1996) and Walters and Bailey (1998) suggested high-grade metamorphism
increased the grain size of ore minerals of the Broken Hill deposit and led to localised
metasomatic remobilisation. A cataclastic texture is common in most of sulphide rocks of
the Broken Hill orebodies and it is characterised by angular to rounded clasts of wall rocks
including metapelite, metapsammite, quartz, plumbian orthoclase, garnetite, garnet
quartzite, quartz-gahnite rocks (Blampain & Plimer 2006; Plimer 1984).
1.6.1 Partial melting
1.6.1.1 Orebody
There are arguments as to whether partial melting has any role in sulphide ore
formation of the Broken Hill deposit. The concept of partial melting of massive sulphide
deposits was developed by Frost, Mavrogenes and Tomkins (2002), Lawrence (1967),
Tomkins, Pattison and Frost (2007) and Vokes (1971). Sparks and Mavrogenes (2005)
argued that presence of inclusions of sulphide minerals in garnet from garnetite indicate
the partial melting of the enclosing garnet. However, Spry, Plimer and Teale (2008)
claimed that this is not correct because P-T conditions of the Curnamona Craton (upper
greenschist-lower amphibolite) were too low for this melting. Moreover, mineralogical
observations show that some of the sulphide inclusions formed in open system such as
fractures of quartz and garnet or along the boundaries. Spry, Plimer and Teale (2008) argue
that the sulphide inclusions were derived from hydrothermal processes during retrograde
metamorphism rather than partial melting.
1.6.1.2 Wall rocks
Wall rocks of the Broken Hill orebodies comprise intermittent garnetite and quartz
garnetite. Mavrogenes et al. (2004) argue that enrichment of Mn in garnetite and quartz |
ADE | 20 Chapter 1-Regional and Local Geology
garnetite are associated with partial melting. They suggested the reaction of Mn-rich
sphalerite on wall rocks as the reason for the increase in Mn content in the wall rocks of
Broken Hill deposit. However, although such a process consumes Mn, in nature, sphalerite
does not have such a high Mn content hence there is problem of mass balance.
Furthermore, Mn rims have formed during the low metamorphic grade Delamerian
Orogeny when P-T conditions were far too low for sulphide rock melting (Plimer 2006b).
Another problem for interpretation by Mavrogenes et al. (2004) is that Mn-rich garnetite
and garnet quartzite are located near the Pb-rich ores of the Broken Hill deposit rather than
Zn-rich ores (problem of distribution of garnetite; Spry, Plimer & Teale 2008).
Spry, Plimer and Teale (2008) suggest exhalation and inhalation effects at or near
the sea floor as the reason of the Fe- and Mn-rich wall rocks of Broken Hill. Garnet
quartzites of the Western Mineralisation are significantly more within the structural
hanging wall rather than the structural footwall and they are interlayered with major
unaltered pelitic and psammopelitic units. The hanging wall of the Western Mineralisation
also shows a range of weak to intense alteration.
1.6.2 The Eastern Mineralisation
The Eastern Mineralisation is characterised by quartz-orange garnet17-hedenbergite,
cataclastic sulphide veins (e.g. quartz-galena ± chalcopyrite-sphalerite veins) and bleached
clasts of pelite and sulphide veinlets (Blampain & Plimer 2006; Plimer 2006a). The major
differences of the Eastern Mineralisation with the west of the lease area are the presence of
stratigraphic units of magnetite-bearing amphibolite, Potosi Gneiss and metasediments and
the absence of the stratigraphic marker Unit 4.6 and its BIF (Blampain & Plimer 2006).
1.6.3 The Western Mineralisation
The Western Mineralisation located in Unit 4.7 and its basal unit, Potosi Gneiss
No.1 (Figure 1.5) is equivalent to stratigraphic spotted psammopelite. The orebody was
formed within syn-metamorphic quartz veins and it is characterised as stringers of
sulphide, disseminated ore and remobilised sulphide ore (Plimer et al. 2003). The
remobilisation is a significant geological factor of the Western Mineralisation that should
17 They are mainly spessartine (Sparks & Mavrogenes 2005) |
ADE | 22 Chapter 1-Regional and Local Geology
The Western Mineralisation comprises three stratabound sequences (Leyh 2000;
Spry, Plimer & Teale 2008), a downdip extensions of A Lode with high abundant
spessartine18 ± rhodonite19 unit equivalent to garnet quartzite (Haydon & McConachy
1987), a hedenbergite-rich unit equivalent to B Lode (Gentle 1968) and a quartz-gahnite-
bearing sequence equivalent to C Lode. In this study, the silicate minerals and rock types
of the three stratabound units where intersected the sulphide rocks are quantified to
understand whether there is a quantitative relationship between the silicate minerals
and mineralisation. There is a distinct gradation of metapelite and feldspathic
metapsammopelite into garnet-bearing metapelite and garnetite in the stratigraphic
footwall of both C Lode and the Western Mineralisation.
The Western Mineralisation is terminated downdip by the Globe Vauxhall Shear
Zone which has a high angle to bedding. It consists of several closely-spaced shear zones
within a wide-retrogressed zone (Figure 1.5). The Centenary Mineralisation is located
structurally beneath the Globe Vauxhall Shear Zone and may be equivalent to the Western
Mineralisation or may be in the upper segment of Unit 4.5. The Centenary Mineralisation
zone appears mostly within blue quartz lode, garnet gangue and calc-silicate gangue in
addition to garnet quartzite.
1.6.4 Mineral chemistry and alteration
Mineral chemistry of the Broken Hill ore deposit differs from other known base
metal massive sulphide ore deposits because of deformation, remobilisation and chemical
fractionation during the Olarian and Delamerian Orogenies. Some of the fundamental
differences are:
1. The lack of pyrite (FeS ) in the Broken Hill orebodies and yet the abundance of
2
pyrite in the Broken Hill Block, e.g. Thackaringa Group (Plimer, 1977) indicate
that the Broken Hill orebody has originally a low S content and that S could not
have been released during granulite facies metamorphism,
2. The lack of barite (BaSO ) and other sulphates in most Broken Hill-type deposits
4
(e.g. Pinnacles) and the presence of hyalophane-celsian rocks in the Broken Hill
Domain stratigraphically equivalent to barite rocks in the Olary Domain,
18 Mn Al (SiO )
3 2 4 3
19 MnSiO often contains minor Ca, Fe and Mg
3 |
ADE | Chapter 1-Regional and Local Geology 23
3. The low Cu and Au content of the massive sulphide rocks with slight enrichment in
the garnet-rich rocks associated with 3 Lens. These Cu and Au enrichments are
associated with enrichment in As, Ag, Mo and W,
4. The high Mn, Ca, halide (F, I and Br) and phosphorus content (Plimer 1984), and
5. The high U content of immediate wall rocks (Plimer 1979).
The stratigraphically lowermost of the Broken Hill orebodies is characterised by
enrichment of Zn, Cu, Bi, P, Ni and Ca and depletion of Pb, Ag, Mn, Sb, As and F in
comparison with stratigraphic uppermost of the Broken Hill (Haydon & McConachy 1987;
Johnson & Klingner 1975; Plimer 1979; Spry, Plimer & Teale 2008). Hydrothermal
alteration of oxides and Pb, Zn, U, S and Fe increase toward the Broken Hill orebodies are
characterised by depletion in Na O, CaO, Sr, MgO and enrichment in SiO , K O, Rb,
2 2 2
MnO, Pb, Zn, U, S, P O and total Fe and TiO (Plimer 1979). Moreover, during this
2 5 2
Rb
hydrothermal alteration, the ratio of increased from less than 500 m from the orebodies
Sr
to 30-100 m at the orebodies (Plimer 1979). In this study, the spatial distribution and
variation of 10 elements of the Western Mineralisation are plotted and interpreted.
There has been little isotopic work undertaken on the sulphide masses at
Broken Hill (apart from Pb and S) and previous work has been geologically unconstrained.
The literature does not show that any authors have ever considered that the formation of
the Broken Hill orebody may have been a long-lived geothermal process whereby earlier
sulphide masses were replaced by later sulphide masses, as is seen in Kuroko-type
deposits. This suggests that rather than zonation from C Lode to 3 Lens viewed through the
Principle of Superposition, earlier Zinc orebodies (e.g. A Lode) may be replaced by later
Pb-Mn rich fluids.
1.7 Unsolved problems and research questions of the Broken Hill orebodies
The development of mining operations and the evolution of knowledge, scientific
matters concerning the Broken Hill orebody will always be contentious. The greatest
geological problem at Broken Hill is that, although there have been generations of mine
and exploration geologists, there has been no consistent system of rock nomenclature and
qualitative core logging and hence there is no detailed stratigraphic, structural, geophysical |
ADE | 24 Chapter 1-Regional and Local Geology
and geochemical model of the Broken Hill orebody. Core logging has been non-numerical
and subjective and it has suffered from generations of geologists qualitatively logging core
differently.
To gain an understanding of the stratigraphic setting of the Broken Hill ore deposit,
there needs to be a consistent coherent logging system for Broken Hill and, although the
Southern Operations (NBHC, Zinc Corporation, Pasminco and now Perilya), Broken Hill
South, CBH Resources, North Broken Hill and the Geological Survey of NSW have all had
logging systems and nomenclature, these systems are all different are not able to be used
for correlation. This research project was undertaken on the Western Mineralisation on
CML7 in order to address the following unsolved problems:
a. The Western Mineralisation has been documented using conventional qualitative
core logging and previous collected samples [geological, geotechnical, engineering
and AAS20, wet chemical, ICP-OES21 assaying, XRF22, EMPA23 and ICP-MS24 ] of
oriented and non-oriented core. No geophysical logging has been undertaken and
the previous quantitative information was not well integrated so that it could be
used simply for a variety of statistical analysis and thus the maximum information
has not been extracted from core.
b. Conventional qualitative core logging is separated from mineral chemistry and
geophysical data which is commonly undertaken by mine geologists rather than
research scientists. Logging is constrained to a number of rock types (metapelite,
metapsammopelite, metapsammite, BIF, amphibolite, garnet amphibolite, Potosi
Gneiss, felsic gneiss and pegmatite), these rock types are not related to mineral
chemistry or whole rock geochemistry (hence subtle pre-metamorphic alteration is
ignored) and assays are only undertaken where visible sulphides are present.
Assays are on a different scale to lithological logs and they are added as a flitch
later with MicromineTM and VulcanTM models for mining. Such a diversity of
information on different scales cannot be evaluated simultaneously. However, the
20 Atomic Absorption Spectrophotometry
21 Inductively Coupled Plasma-Optical Emission Spectrometry
22 X-Ray Fluorescence
23 Electron Microprobe Analysis
24 Inductively Coupled Plasma Mass Spectrometry |
ADE | Chapter 1-Regional and Local Geology 25
processes undertaken at the Rasp Mine are standard industry practice for the
economic mining of sulphide orebodies and do not allow the easy use of the data
for performing different statistical methods and providing quantitative core log
diagrams and bar diagrams.
c. Geostatistical methods are used at the Rasp Mine to validate assay data, to test the
veracity of data, to construct economic mining models and to calculate resources
and reserves but the quantitative core log data can be evaluated using a variety of
classic bivariate and multivariate statistical methods to improve quality of
interpretation. There have been a number of due diligence and ore resource/reserve
studies undertaken by CBH Resources Ltd and their consultants. However, the
answers sought in such studies were commercial and not scientific hence were
unable to address some of the unsolved problems above.
d. There is a lack of a quantitative understanding of the spatial relationship and
structural continuity of geological, geochemical and geophysical features of the
Western Mineralisation. The spatial and structural characteristics of the Broken Hill
ore deposit including the Western Mineralisation resulted from a multiphase
evolving geothermal system that has undergone a superimposition of a multiphase
tectonism and deformation. The events created hydrothermal alteration, induced
sulphide flow structures in silicate rocks, remobilised sulphide minerals, distilled
some elements from the sulphide rocks and changed the orebody morphology.
Spatial models of quantitative core log data of the Western Mineralisation can
address the following unknown issues:
1. The nature and morphology of geochemical zonation haloes,
2. Spatial texture of sulphide minerals,
3. Classification of geochemical anomalies,
4. Separation of threshold values from background and anomalous levels for
different elements,
5. Identification of geochemical and mineralogical zonation patterns,
6. Spatial anisotropy, |
ADE | 26 Chapter 1-Regional and Local Geology
7. Spatial pattern of magnetic susceptibility,
8. Detection of geochemical pathfinders of Pb and Zn, and
9. Identification of geochemical zonation sequence in different directions of
the orebody.
During this study, there were a number of validation questions that were considered
in the research. Some of the questions are:
1. Were the various type of data sets (geology, geochemistry and geophysics)
used accurate and can results from different statistical methods be
compared?
2. How were samples collected?
3. What sample intervals were used and how were these samples related to
each other?
4. Were different data sets cross checked, re-analysed and validated?
5. How was qualitative geological data integrated with quantitative
geochemical data, bearing in mind that these data sets were collected on
different scales?
6. Why should such data be integrated?
The aim of this project is to analyse and interpret these old data sets (assay) and
integrate data sets generated during this study (geophysics, lithology, mineralogy and
sulphide textures) by univariate, bivariate, multivariate analysis and geostatistical methods
to evaluate the degree of uncertainty of the interpretations. As the Rasp Mine is in the
planning stage, these interpretations will be timely as ore resource calculations show that
there is high, medium and low grade ore yet there is no predictive methodology for grade. |
ADE | CHAPTER 2
Quantitative Core Logging in the Western Mineralisation
2.1 Introduction
There has been more than 100 years of drilling at CML7 in Broken Hill on seven
different grid systems in feet, fathoms and metres. Most cores are not preserved, there are
no logs of many of the holes from the 19th and early 20th Centuries and only some of this
data is preserved as assay plots on maps. It is only in the last 40 years that core has been
stored in dedicated core facilities. On CML7, none of the core from drilling by previous
CML7 mining companies is preserved (BHP, Sulphide Corporation, Block 10 Silver
Mining Company, BHP Block 14 Company, North Broken Hill, South Broken Hill and
Minerals Mining and Metallurgy) and there is only partial preservation of core from
Normandy Mining Ltd’s exploration activities. CML7 was acquired by Redfire Resources
N.L. (now CBH Resources Ltd) in 2001 and all drill core since that time is preserved.
Another 4,500 m of underground drilling of the Western Mineralisation from the Rasp
Mine decline is taking place at present. Fanned holes drilled westerly are designed to infill
areas where drilling used in this study was sparse. This new drilling is to convert resources
to reserves to make the project more bankable and to more closely define the Western
Mineralisation for stope design.
Assay values from underground drilling from the 1000', 1150', 1250' and 1480'
levels by BH South Ltd in the 1950s and 1960s are plotted but no duplicate samples
remain, no core remains and the core was not geologically mapped. No grade control
underground drill core is preserved and some of these data sets are plotted onto old maps.
Because the 127 years of continuous mining on CML7 has been by numerous operators
with different economic constraints, the only data on the Western Mineralisation used in
this study was that of CBH Resources Ltd because core is available for validation and
cross checking. Nevertheless, there have been 4 geologists (Pascal Blampain, John Collier,
Ian Plimer, Catherine Errock) who have logged CBH Resources Ltd core and relogged the
Normandy Mining Ltd core over the last decade. There was an attempt by Blampain to
standardise nomenclature and logging techniques hence there is a high degree of internal
consistency for qualitative logging. |
ADE | 28 Chapter 2-Quantitative Core Logging
The Western Mineralisation sample cores were not logged by CBH Resources Ltd
for ore texture, ore modal mineralogy, ore structure and magnetic susceptibility but the
data collected in this study was integrated with metre-by-metre assay data, mineralogical
and lithological logs of CBH Resources Ltd.
This chapter addresses the uncertainties associated with conventional core logging
(as shown by the Western Mineralisation) and introduces a numerical approach for the
quantification of visual estimates of minerals, rock types and textures and the relationship
of this data to mine assay data. These relationships are shown on Western Mineralisation
drill collar maps (Figure 2.1) and 3D visualisation of sample locations (Figure 2.2).
Satellite imagery of the Western Mineralisation has been provided in supplementary file to
this thesis.
2.2 Maps
2.2.1 Drill collar locations
The 54 CBH Resources Ltd surface drill holes into the Western Mineralisation are
shown on Figure 2.1. In Figure 2.1, four drill holes 3230, 3231, 3232 and 3233 are related
to Normandy Mining Ltd’s previous exploration activities and other drill holes are related
to the Western Mineralisation’s exploration activities. The drill hole numbers of the
Western Mineralisation were named by a prefix "WMDD1" (e.g. WMDD4001) but in this
thesis, only the drill hole numbers (e.g. 4001) are used in figures and text.
Surface drill locations were 1,500.7 m in the south (drill hole 4052), 2,562.3 m in
the north (drill hole 4034), and 9,249.9 m in the west (drill hole 4042) to 9,531.88 m in the
east (drill hole 4054). Horizontal angles relative to strike were calculated on a local CBH
Resources Ltd graticule system hence true north is different from grid north. Therefore,
azimuth of the Broken Hill Mine toward northing is slightly different from the
geographical azimuth at the Broken Hill district. There are now 31 CBH Resources Ltd
underground holes drilled from the Rasp Decline into the Western Mineralisation. Some of
the underground drill cores have assay values and only their assay values have been used
in this study.
1 Western Mineralisation Diamond Drill core |
ADE | Chapter 2-Quantitative Core Logging 29
Based on the existing standard collar location map of CBH Resources Ltd, the
northing and easting data were converted to local territory and for this purpose values of
1,000 and 10,000 were added to the original values of the north and east of the Western
Mineralisation respectively. Moreover, the easting values are increasing from the west
toward the east in the collar map. In this collar map, the contour lines show the relative
surface level of the drill holes in metres. The perimeter of the collar map is 2900 m and the
surface area corresponds to 385,000 m2.
There is a number of surface drill holes that were drilled from the same collar but
with different dips and they produced a series of fanned holes. In Figure 2.1, this can be
seen in several cases where a single point is attributed to different drill holes. Furthermore,
Figure 2.1 shows that most drill holes in the Western Mineralisation drilled along a strike
of approximately 15° toward NE from the local north coordinate of the CBH Resources
Ltd.
The surface drill holes were drilled in an irregular pattern (Figure 2.1). This
arrangement of drill sites was chosen because of logistical problems such as vicinity of the
Western Mineralisation to the City of Broken Hill and interference with existing buildings,
roads and railways (see the satellite imagery of the collar locations of the surface drill
holes in supplementary files to this thesis). Moreover, the drill hole numbers in the collar
map are not in a regular order and were named by CBH Resources Ltd based on the
chronological order of drilling. |
ADE | Chapter 2-Quantitative Core Logging 31
2.2.2 Three-dimensional visualisation of the analysed sample locations
Sample locations of the Western Mineralisation were plotted on a 3D grid (Figure
2.2). The 3D sample locations shown are for the drill cores selected for both assaying and
detailed geological examination in this study. The 3D sample locations are provided at a
scale of 1:6,000 m for all directions (northing, easting and elevation). In Figure 2.2, the
elevation is increasing upwards in contrast with depth. The top elevation (10,222.33 m)
shows the start of sampled drill hole 4032 and the bottom elevation (9,691.69 m) is related
to the end of sulphide mineralised sample in drill hole 4052. Total depth between start and
end of sampled drill hole 4032 and 4052 respectively is 530.64 m. The local elevation
values were summed at 10,000 in accord with the base map of the CBH Resources Ltd for
the Western Mineralisation.
The sampled drill locations were gradually deeper towards the south-western corner
of the 3D map (Figure 2.2). The magnetic azimuth calculated from the local grid azimuth
is plus 43°. The depth of surface drill holes varies from 141 m to 698.8 m and the three
drill holes in Normandy Mining (3232, 3231 and 3230) were deeper with a hole end depth
of 830.8 m. However, there are no assay values available beyond 530 m depth in any of the
drill holes.
The length of underground drill holes ranges from 107.3 m to 402.4 m. The dip of
surface drill holes varies from -47° to -88° and the dip of underground drill holes ranges
between +10 and -90. It should be noted that term dip, in this section is a mining concept.
Negative dip indicates moving down from the horizontal surface level and positive dip
means moving up towards the horizontal surface level. |
ADE | Chapter 2-Quantitative Core Logging 33
2.3 Geological information resulted from core logging in the Western Mineralisation
2.3.1 The conventional core logging
In mines, a set of quantitative data and descriptive geological information is
provided for each drill hole. The quantitative data usually consists of the core size, azimuth
and dip of drill holes, collar location of drill holes and depth of the major rock type, rock
quality designation (RQD), recovery percentage, rock strength, orientation reliability and
fracture density as well as specific gravity. Descriptive geological information is
commonly provided for petrography and dominant sulphide and silicate minerals, texture,
alteration or weathering, deformation and metamorphic minerals from macroscopic visual
examination.
The recovery percentage of the core intervals is calculated by measuring the total
length of core (using the tape measure) that has been marked up as percentage over the one
metre interval. An empirical method for estimating core recovery is RQD and it is
calculated by summation of the length of all pieces of cores if they are individually greater
than a length of 100 mm. For example, if the core is recovered in pieces that are all longer
than 100 mm, the RQD is 100 % whereas if some pieces are shorter than 100 mm, the
RQD will decrease. The RQD value provides general information regarding the degree of
jointing and faulting in the rock mass. It is the procedures that are utilised by CBH
Resources Ltd as their conventional qualitative core logging of the Western Mineralisation
(Table 2.1). |
ADE | Chapter 2-Quantitative Core Logging 35
4. Absence of assay values and geophysical results (e.g. magnetic susceptibility and
specific gravity) in the conventional core logs for evaluation of internal consistency
and coherency.
The geological information of core samples of the Western Mineralisation has not
been based on variation of the sulphide mineralisation, but is based on variation of rock
type intervals. In this case, attribution of a large length of core (e.g. 20 m) to one rock type
may be useful for description of the non-mineralised core intervals (barren units) and low
grade sulphide rocks, but in the rich-sulphide rocks (economic ore zones or productive
zones) cause loss of quality and quantity of the geological information. Moreover, it is not
clear where the reported geological information over a long length of core relates to any
particular part of the core. Core investigation in equal length core sections increases the
number of investigated geological samples and improves the quality of geological
observations.
2.3.2 The quantitative core logging
The aim of this project is to construct comprehensive and quantitative geological
information for 1,928 m of core samples of 54 surface drill holes to supplement the
qualitative logs of CBH Resources Ltd. The first step was to identify major geological
parameters for providing a quantitative core logging. The following geological information
was selected for this study:
1. Silicate mineral assemblages: green feldspar, pink garnet2, orange garnet, red
garnet3, hedenbergite, rhodonite, white quartz, gahnite and blue quartz.
2. Sulphide minerals assemblages: galena, sphalerite, chalcopyrite, pyrrhotite, pyrite
and arsenopyrite, and
3. Rock types: dolerite, garnet quartzite (quartz garnetite), quartzite lode, pegmatite,
metapelite, metapsammite, metapsammopelite and blue quartz lode
Blue quartz lode
The term blue quartz lode (blue quartz ± gahnite ± minor garnet) has been used in
most internal mining reports of the Broken Hill Mine and the conventional core logs of the
Western Mineralisation as a rock type. This rock type appears broadly distributed in the
2 They are mainly pink almandine [Fe Al (SiO ) ]-pyrope [(Mg Al (SiO ) ]
3 2 4 3 3 2 4 3
3 They are mainly spessartine (Sparks & Mavrogenes 2005) |
ADE | 36 Chapter 2-Quantitative Core Logging
stratigraphic package of the Broken Hill orebodies. Blue quartz lode has also been used in
some papers (e.g. Morland & Webster, 1998; Plimer, 2006b). In this study, in a few core
samples, the major host rock was not blue quartz lode but the samples containing a small
volume percentage of blue quartz. In these cases, it was accounted for as a mineral.
Therefore, the volume percentage of blue quartz was quantified both for rock group (blue
quartz lode) and for mineral group (blue quartz). Nevertheless, in most cases, the
abundance of the blue quartz lode was estimated similar to the blue quartz in mineral
group.
2.3.3 Quantification of minerals and rock types
In the Western Mineralisation, each one-metre core section was examined in
smaller parts based on variation of rock types and mineral content. The average volume
percent of the individual minerals and rock types were reported for each one-metre core
sample. For quantification of minerals and rock types, they were classified into the three
following groups based on their grain sizes, distribution and texture.
2.3.3.1 Quantification of minerals and rock types in group one
Group one includes all the investigated rock types of this study and some coarse
grain silicate minerals (e.g. green feldspar, white quartz, blue quartz and garnets)
and sulphide minerals (e.g. galena, sphalerite and pyrrhotite) in high grade sulphide
mineralised samples (Figure 2.3). Major textures of sulphide minerals in this group
comprise: massive, brecciated, high grade veins, stringer and network.
Figure 2.3: Coarse sulphide minerals including galena (grey), sphalerite (black), chalcopyrite
(yellow) and white quartz in a cataclastic massive sulphide rock of the Western
Mineralisation (drill hole 4033 at 353.2 m). The structural position of cataclastic ore is
unknown and may well remain so until further underground mining commences. |
ADE | Chapter 2-Quantitative Core Logging 37
In Figure 2.3, the sulphide and silicate minerals have entirely occupied the clear
part of each metre split core and rock type. Minerals are visually identifiable by their high
continuity and massive textures. In this group, the amount of minerals ranges
from 10 to 65 vol. % 4. The volume percent of rocks was estimated using a tape measure.
However, in some samples, any obvious boundary between two different rock types was
equivocal because the boundary is gradational (e.g. metapsammopelite as a transitional
rock between metapelite and metapsammite). Furthermore, the volume percentage of rock
types is biased when the core axis to bedding angle approaches parallelism. In highly
mineralised samples, the estimation of sulphide mineral content is problematic because an
intergrowth of several sulphide and silicate minerals masks the mineral lustre. In order to
reduce this error, the cores were viewed both wet and dry.
2.3.3.2 Quantification of minerals in group two
Group two consists of some silicate minerals including pink garnet, red garnet,
orange garnet, hedenbergite, rhodonite and gahnite. Minerals of this group are
characterised by a regular dissemination of grains, equal crystal size and distinct colour
along one metre core intervals (e.g. Figure 2.4). The quantifying process in this group was
performed based on mineralogical comparison charts of Figure 2.5 (Terry & Chilingar
1955). In this group, the amount of minerals varies between 5 and 50 vol %.
Figure 2.4: Distribution of pink garnet (red crystals) within garnet quartzite in drill hole
4003, between 287.6 m and 287.75 m.
4 Volume percent |
ADE | Chapter 2-Quantitative Core Logging 39
In this group, sulphide minerals comprise less than 5 vol. % of the Western
Mineralisation. These sulphide minerals are not economically significant. However,
estimation of minerals of this group produce great information about ore mineral zonation
pattern that sometimes functions as pathfinder for tracking mineralised zones. The spatial
structure of the ore mineral zonation pattern depends on distinguishing both parts of low
and high grade sulphide mineralised zones of the Western Mineralisation.
2.3.4 Some effective parameters and considerations in the quantitative core logging
Estimating volume percentage of minerals and rock types is an empirical method
that depends on the geological, mineralogical and human parameters. Some of the
parameters have been outlined in Table 2.2.
Table 2.2: Effective parameters in quality of the quantitative core logging.
Mineralogical Equal or unequal distribution of minerals, grain size, texture, colour and the
parameters proportions of minerals in mineral mixtures.
These parameters may have indirectly significant effects on the quantification
Human process. It can be improved by development of practical core logging,
parameters concentration, identification of local rock types and quick classification of
samples based on the known local rock types.
Geological Geological complexity such as the number of minerals, rock types and textures
parameters present in a core sample.
Based on the practical experience of this study, some considerations (Table 2.3)
have shown to be a useful for minimizing uncertainties during quantification of minerals
and rock types. |
ADE | 40 Chapter 2-Quantitative Core Logging
Table 2.3: Useful considerations for quantification of minerals and rock types.
The quantitative core investigation can be improved if carried out by one person for the sample
set in order to get a coherent core logging.
Previous experience of conventional core logging helps to improve speed of the quantitative core
logging.
Using split cores for visual examination helps to identify minerals better and improve the
quantitative estimate of minerals and rock types.
Paying attention to information recorded in the conventional core logs improves quantitative core
logging for the samples.
Quantification of galena and sphalerite in the presence of Pb and Zn assays respectively may
produce a bias in visual grade estimation and visual modal mineralogical estimation.
Identification of important physical properties of each mineral (e.g. shape, size, lustre, colour,
textures, associated minerals, textural continuity, distribution, cleavage, crystal form and
enveloping rocks) increases the reliability of visual recognition and estimation of minerals.
Using daylight and a hand lens.
It is essential to examine sulphide minerals in dry core because wet core exacerbates sulphide
mineral lustre and affects modal estimation.
Silicate minerals and rock types of wet core are more consistent.
Magnetic susceptibility should be measured on dry samples.
It is necessary to perform visual estimation of minerals and rocks for a set of similar samples
during a continuous period of time; otherwise, there is the possibility of losing coherency.
Random checking of the quantified minerals or rock types in previous samples in order to
increase accuracy, precision and reduce the risk of human error.
If the results of assay data are available, the estimated galena and sphalerite can be reconciled
with Pb and Zn assays respectively.
2.3.5 Quantification of textures for sulphide minerals of the Western Mineralisation
There are some textural descriptions in the conventional core logging of the
Western Mineralisation by CBH Resources Ltd but the conventional core logs do not show
clearly in which part of the drill cores they were observed. In the Western Mineralisation,
dominant textures within rich sulphide samples were network and massive (suggesting
remobilisation) whereas in low grade sulphide rocks, the major textures were disseminated,
vein and laminated. |
ADE | Chapter 2-Quantitative Core Logging 41
The stringer texture was present in both high grade and low grade sulphide
samples. An indicator was used for quantification of the sulphide textures. In this method,
a value of one was given to occurrence of each texture that was seen within each metre
core. Based on this, spatial probability of occurrence of textures was calculated in Chapters
6 and plotted in Chapter 8. Identification of textures may be indirectly helpful for
prediction of dilution effects in mining when, for example, a high grade sulphide zone has
unexpected textural changes that will affect grade control.
One good example of application of the indicator in geological study is related to
De Geoffroy and Wignall (1972). They quantified visually rock types, minerals and
structural geology in several porphyry deposits of Cu and Mo in the Cordilleran Belt in
order to identify their similarities and differences. They processed their results with
character analysis but character analysis does not indicate the volume percentage of
minerals or rock types.
2.3.6 Previous visual quantification methods in geosciences
Visual quantification of the Western Mineralisation is not limited to this study and
core samples were quantified visually for recovery percent, RQD and fracture density for
which all the data is purely empirical. For example, in the Western Mineralisation, fracture
density quantification, the quantity of bedding and jointing was measured over one metre
intervals.
2.4 Geochemical data of the Western Mineralisation
In the Western Mineralisation, the selected intervals of drill cores for assaying
comprise sulphide mineralised rocks, poorly mineralised zones and to some extent their
enclosing unmineralised host rocks. The core samples were cut with a diamond saw along
their long axis and one half of the core was retained in the core tray and other parts of the
split cores of different lengths (between 0.5 and 1.75 m) were sent for assaying. In the
Western Mineralisation, the surface drill core samples of CBH Resources Ltd were assayed
by ICP-OES for 10 elements [Pb %, Zn %, Fe %, S %, Cu %, Ag (ppm), Cd (ppm), Sb
(ppm), Bi (ppm), As (ppm)] and the underground samples assayed for 8 elements [Pb %,
Zn %, Fe %, S %, Ag (ppm), As (ppm), Sb (ppm) and Bi (ppm)]. |
ADE | 42 Chapter 2-Quantitative Core Logging
2.4.1 Sample size reconciliation
Western Mineralisation drill core was HQ (63.5 mm5) and NQ size (47.6 mm9),
core was halved for assaying and re-sampling and metallurgical bench testing left quarter
core. One large diameter PQ core (85 mm9) was drilled for metallurgical testing and the
entire core was used. With coarse grained sulphides, stringers and transgressive zones of
sulphides, it is quite probable that each half core interval would give different assay data.
In order to reduce such uncertainties and biased geochemical data, equivalent sample
support is required otherwise it has to be shown that there is no relationship between the
element concentration, core diameter and the length of core analysed.
2.4.1.1 Equivalent core diameter
Different core diameters are used for surface and underground drilling. Western
Mineralisation surface holes started as HQ until the base of weathering (10-45 m) allowing
tighter directional control. Once in hard rock, core diameter was reduced to NQ. Larger
core diameter has better recovery however it is more expensive to drill and more difficult
to handle. Variable core diameter in the evaluation of the Western Mineralisation may
increase uncertainties. All sulphide mineralised core samples of this study were selected
from NQ parts of the surface drill holes apart from holes 4001, 4002, 4031 and 4048 that
were HQ size. The 23 underground Western Mineralisation drill holes from drill cuddies6
in the Rasp Decline were LTK60 size (45.2 mm9).
If different core diameters are used for grade estimation, it is possible that the grade
may not be accurate. Equivalent sample sizes are essential for accurate statistical analysis
and comparison of element concentration because increase of sample volume (V) leads to
decrease of variance (S2, where S is standard deviation) of element concentration.
Hazen’s study in 1967 (cited in Wellmer 1998, p.24) based on experimental drilling
project undertaken at the Climax molybdenite mine near Denver (Colorado) suggested an
empirical formula [Equation (2.1)] that shows the sample volume-variance product
should be constant within a given orebody.
S2 V Constant (2.1)
5 Core (inside) diameter
6A mining term for a storage area or drill site off the main drive or decline |
ADE | Chapter 2-Quantitative Core Logging 43
According to Equation (2.1), two samples with equal support collected in a close
distance from one type of mineralisation have similar variance for a measured variable
(e.g. element concentration). Koch and Link’s study in 1970 (cited in Wellmer 1998, p.24)
experimented the sample volume-variance [Equation (2.1)] for several core diameters (BQ
to HQ) and they found that the variances of element concentration reduces with a much
smaller proportion relative to degree of increase in their sample volumes.
In nature, the Hazen’s equation may not be accurate if the distance between two
samples increases because it is possible the average trend of mineralisation and grade
change rapidly. Although the Hazen’s equation for different distance of samples does not
strictly occur in nature and specific geological properties impact on this approach, the
sample volume-variance relationship [Equation (2.1)] is used for comparison of sample
series with different support and possible correction (Wellmer 1998). In the Western
Mineralisation, Hazen’s equation was used for conversion of the HQ and LTK60 core
diameters to the NQ core diameter at the same drilling position. In this case, there were
fewer problems of a large distance between two samples that caused the increase in
uncertainty of grade variation of mineralisation.
The Western Mineralisation is not a vein or disseminated mineralisation but it
is a stratabound ore zone which is geometric, textural and grade variation. However, core
logging shows that such variations are rare over a few centimetres around drill holes.
Equation (2.2) is used for comparing sample volume of V with concentration variance of
1
S2
and sample volume of V with concentration variance of
S2
. It is a consequent of
1 2 2
Hazen’s equation that suggests the variance of concentration inversely changes with
proportion of the sample volume but the product of volume-variance is constant
approximately (Wellmer 1998, p.24).
S2 V S2 V
(2.2)
1 1 2 2
All assayed parts of HQ and LKT60 drill holes were considered as a long core
sample for converting to equivalent length of NQ core sample. In Table 2.4, Equations
(2.3) and (2.4) are used for calculation of standard deviation of an equivalent NQ core
sample from standard deviation of the HQ and the LTK60 core size. It should be noted that |
ADE | 44 Chapter 2-Quantitative Core Logging
the length of equivalent NQ drill cores is the same for the initial HQ and LTK60 drill
cores. Therefore, the length (l) is eliminated in methodical conversion of Table 2.4.
Table 2.4: Calculation of standard deviation of an equivalent NQ core sample from standard
deviation of the HQ and the LTK60 core size.
πr2 l V
d2 V
π l Split Core
4 2
S2 V S2 V
HQ HQ NQ NQ
V d 63.5
S S HQ S HQ S
NQ HQ V HQ d HQ 47.6
NQ NQ
S 1.33403S (2.3)
NQ HQ
V d 45.2
S S LTK60 S LTK60 S (2.4)
NQ LTK60 V LTK60 d LTK60 47.6
NQ NQ
where r = Radius, l = Length of a core sample, d = Core diameter and V = Core volume
In theory, the mean grade of HQ and LTK60 samples should be considered the
same for NQ samples by conversion using the volume-variance comparison line (VVC
line; Figure 2.7; Wellmer 1998). If the mean of concentration of the two samples is not the
same, according to Wellmer (1998), it must be assumed a systematic error and bias in drill
samples. In this case, construction of the VVC line is not appropriate for this conversion or
sample correction. |
ADE | Chapter 2-Quantitative Core Logging 45
Figure 2.7: The VVC line for comparing the variances of an element concentration
between HQ and NQ samples support (redrawn from Wellmer 1998). S and S are
x y
standard deviations of concentrations for NQ and HQ samples support respectively.
X and Y are the mean concentrations for NQ and HQ samples support respectively.
In Figure 2.7, the mean concentration Yof the HQ samples was considered equal
to the mean concentration Xof the NQ samples. The two mean points of X and Yin
horizontal and vertical axis respectively determine one point (1) of the VVC line. The
nature and dimension of the standard deviation are the same as the mean value and can be
added or subtracted to a mean value. Two points of S and S can be calculated from the
x y
standard deviation of concentration of the NQ drill core and its equivalent standard
deviation of the HQ drill core respectively using Equations (2.5) and (2.6).
S
S X NQ (2.5)
x 2
S
S Y HQ (2.6)
y 2 |
ADE | 46 Chapter 2-Quantitative Core Logging
The VVC line can be constructed by two points 1 and 2 (on red line in Figure 2.7)
resulting from intersections (dashed black lines) of two points(X,Y)and (S ,S ).This
x y
process is performed in Excel to calculate intercept and slope of the VVC line for part of
an assayed drill core. The process of volume-variance comparison may produce a few
negative concentration values for NQ core. The negative data cannot be used for
concentration and the simplest solution is to set them to zero. This is a reasonable solution
provided the negative concentration values are not too many and their magnitude is not too
large. The consequence of the elimination of negative concentration values has led to a
minor decrease in resulting variance and small increase in resulting mean. In the Western
Mineralisation, the small number of HQ drill cores after converting to NQ drill cores
showed a few minor negative values related to low concentration values.
2.4.1.2 Equivalent length of core
According to the theory of volume-variance relationship, the length of a sample has
an indirect relationship with the variance of concentration (grade). It does not mean that
the difference between a few centimetres and a couple of metres core is the same as the
difference of their variances of concentration. The core sections of different length should
not just be integrated in geostatistical estimates and statistical analysis and they need
"regularisation". The regularisation is a term used for compositing assayed core samples to
a permanent interval length (e.g. one metre). The basic regularisation method combines
core samples to equal length of composites and assigns a single concentration to each
composite based on a weighted average of the length of core samples of the original
component concentration (Smith 1999). There are several mathematical methods for
regularisation and compositing of concentration (e.g. Dutter 2003). In this research,
regularisation was performed by the subprogram "Data recomposition" in the Geostatistics
for Windows software package (Dowd & Xu 2006).
2.4.2 Inclination of drill hole
The position of some points of each drill hole along the depth is characterised by
their collar and survey. In the Western Mineralisation, the collar of each point of drill hole
contains the following information:
Easting location with negative signs and the values were summed at 10,000, |
ADE | 48 Chapter 2-Quantitative Core Logging
end of each selected interval core samples. In Analyse program, easting, northing and
elevation of each sample are calculated by Equations (2.7), (2.8) and (2.9) in Table 2.5.
Table 2.5: Formulae used for calculation of the easting, northing and elevation of the samples
of the Western Mineralisation.
Dip 90 Bear
East East (Depth Depth )Sin ( i1 )Sin ( i1) (2.7)
i1 i i1 i I I
Dip 90 Bear
North North (Depth Depth )Sin ( i1 )Cos ( i1) (2.8)
i1 i i1 i I I
90-Dip
RL (Depth Depth )Cos ( i1) (2.9)
i i1 i I
where
180
I 57.295779531 ,
π
i 0 , i i1
In Excel sheet (1), for checking accuracy of the resulting coordinates, the real
distance [Equation (2.10)] and the estimated distance [Equation (2.11)] are both calculated
for each interval core samples along depth (Table 2.6) and their results should be equal to
each other.
Table 2.6: Formulae used for calculation of the real and estimated depth of the samples
in the Western Mineralisation.
Real distance Depth Depth (2.10)
i1 i
Estimated distance (East East )2 (North North )2 (Rl Rl )2 (2.11)
i1 i i1 i i1 i
where i=2 (i.e. the second raw of the Excel sheet).
2.4.3 Recovery percentage
The total length of core is measured using a tape measure and then marked up as a
percentage over one metre core intervals. The resulting percentage is recorded next to
relevant depth interval in core. The recovery percentage should be reviewed to identify the
occurrence of friable and unconsolidated minerals along the core samples as well as to
determine the amount of core loss (during drilling, core washing and core cutting)
particularly in the mineralised part of the cores. |
ADE | Chapter 2-Quantitative Core Logging 49
The result of core loss leads to a systematic downward bias and under-reporting of
the assay values or the ore mineral grade. Core loss will increase significantly if the cores
intersect high grade ore. In the Western Mineralisation, because almost all core samples
comprise consolidated rocks and minerals, more than 99 % of assayed core samples
showed 100 % recovery and the rest of them had 95 % recovery. Hence, in this research
there was no considerable core loss which needed correction.
2.5 Data Compositing
The reported assay values and specific gravity data were recomposited into each
one metre NQ sample and the resulting data was integrated with the quantitative minerals
and rock type data and the measured average magnetic susceptibility (AMS)7 and
maximum magnetic susceptibility (MMS)7. In this research, the following data set
produced:
1. Thirty one geological variables of 1,928 surface core samples comprising 59,768
geological sample values,
2. Three physical parameters (specific gravity, MMS and AMS) of 1,928 surface core
samples comprising 5,784 physical parameter sample values,
3. Ten element concentrations of 1,928 surface core samples comprising 19,280
assayed sample values,
4. Eight element concentrations of 1,287 underground core samples comprising
10,288 assayed sample values, and
5. In total, 77 drill cores in the Western Mineralisation treated for the data
compositing of 95,120 sample values.
In the quantitative core log data, the rows denote the selected core sections
(observations) and the columns show the rock types, sulphide and silicate minerals,
sulphide textures, element concentrations, magnetic susceptibility and specific gravity. The
columns were classified into two of the following independent groups so that each group
adds up to 100 percent:
7 Details of magnetic susceptibility measurements are explained in Section 4.4 |
ADE | 50 Chapter 2-Quantitative Core Logging
1. Group of minerals assemblage: The volume percent of sulphide and silicate
minerals were estimated and their summation was theoretically 100 percent.
However, in many cases, the total was not 100 percent because of the remaining
volume percent of other silicate minerals in the samples (see the data base of this
thesis), and
2. Group of rock types: The volume percent of different rock types plus the volume
percent of sulphide minerals should be 100 percent. Most samples (one metre core
sample) of the Western Mineralisation comprise one or two rock types and a few
core samples consist of more than three rock types (see the data base of this thesis).
The coordinates of northing, easting and elevation for start, middle and end of each
sample interval were calculated for use in variogram analysis and spatial modelling
(Chapters 6 to 8). The quantitative geological, geochemical and geophysical values were
attributed to the middle of each core sample for reasons of consistent mathematical or
statistical treatment of data. The quantitative core log data has been provided in
supplementary files to this thesis.
2.6 Discussion about the conventional core logs 4003, 4004 and 4031
The conventional core logs of the Western Mineralisation describe only lithological
and mineralogical variation with depth and the assay results were usually well much after
the core has been logged and hence could not be used to assist logging. The conventional
core logs would have been combined at a later stage with the assay results using VulcanTM
or MicromineTM at the Rasp Mine and the assay data does not appear on the primary
qualitative core logs (Tables from 2.9 to 2.11). Drill holes 4003 and 4004 were selected as
example of poor quality of information about the sulphide mineralised sections. There are
many similar conventional core logs in the Western Mineralisation that provide little
geological information for sulphide rich segments of their drill holes. Drill hole 4031 was
selected for this section because it is enriched highly in Pb and Zn and it is an example for
good quality conventional core logging of the sulphide intervals. Tables 2.7 and 2.8 shows
codes were used for rock types and minerals respectively in the conventional core logs:
Table 2.7: Lithological codes used in the conventional core logs.
Brec = Brecciated rock Pe = Metapelite Pm = Metapsammopelite
Gq = Garnet quartzite Peg = Pegmatite Ps = Metapsammite
Luq = Blue quartz Lode Pq = Potosi gneiss L = High grade sulphide mineralised lode |
ADE | Chapter 2-Quantitative Core Logging 51
Table 2.8: Mineralogical codes used in the conventional core logs.
Apy = Arsenopyrite Epi = Epidote Gfld = Green feldspar Rho = Rhodonite
Bio = Biotite F = Fluorite Gr = Graphite Ser = Sericite
Blq = Blue quartz Feld = Feldspar Hed = Hedenbergite Sill = Sillimanite
Cc = Calcite Gah =Gahnite Po = Pyrrhotite Sl or Sph = Sphalerite
Chl = Chlorite Gar = Garnet Py = Pyrite Mus = Muscovite
Cpy = Chalcopyrite Gn or Gal = Galena Q = Quartz
2.6.1 The conventional core log 4003
Table 2.9 describes that the content of galena is higher than sphalerite between
344.5 and 364.5 m of this drill hole and the amount of pyrrhotite is higher than pyrite and
chalcopyrite.
Table 2.9: The conventional core log of the sulphide rich segment of drill hole 4003 from
308.1 to 364.5 m. This drill core was logged by Blampain and a new graduate (Collier).
NOTE:
This table is included on page 51 of the print copy of
the thesis held in the University of Adelaide Library. |
ADE | 54 Chapter 2-Quantitative Core Logging
2.7.1 The quantitative core log 4003
In Figures 2.8 and 2.9, the concentrations of Pb, Zn, Cd, S, Fe and to some extent
Cu, sphalerite and galena show similar anomalous peaks and variation trends. Some of the
maximum element concentrations within the sulphide rich segments of this drill hole are:
14.93 % Zn @8 334 m,
5.9 % Pb @ 335 m,
10.17 % Fe @ 361 m, and
153 ppm Cd @ 334 m.
The average concentration of Sb and Bi are 50 ppm but Sb concentration shows
some anomalous peaks between 334 and 336 m and between 339 and 341m.
Some of the anomalous contents of galena and sphalerite are (Figure 2.8):
50 vol. % sphalerite @ 333 m, and
5 vol. % galena @ 335 m.
Some of the anomalous contents of silicate minerals are (Figure 2.9):
20 vol. % hedenbergite @ 347 m,
23 vol. % hedenbergite @ 348 m,
30 vol. % hedenbergite @ 349 m and 350 m, and
15 vol. % orange garnet @ 345 m and 346 m.
In drill hole 4003, the main textures of sulphide minerals are vein, disseminated
and laminated. The abundance of rock types from surface to depth in sequential order are
garnet quartzite, pegmatite, metapelite and metapsammopelite. There is no obvious
relationship between variation of rock types and variation of galena, sphalerite, Pb and Zn
from surface to depth. However, the abundance of pink garnet and gahnite shows a
correlation with the appearance of garnet quartzite. Also, the orange garnet is abundant in
pegmatite.
8 The symbol "@" was used to mark the start of the depth for a core sample |
ADE | 60 Chapter 2-Quantitative Core Logging
2.7.3 The quantitative core log 4031
In Figures 2.12 and 2.13, there are some similarities in variations and peak
anomalies between the following elements and minerals:
Zn, Pb, sphalerite and galena, and
S, Ag and Cd.
In general, variation of all elements shows anomalous peak when approaching
sulphide-bearing intervals. Some of the anomalous concentrations of Zn, Pb and Ag within
this drill hole are (Figures 2.12 and 2.13):
28.10 % Zn @ 305 m,
26.87 % Zn @ 314 m,
26.47 % Zn @ 315 m,
25.07 % Zn @ 316 m,
27.84 % Pb @ 315 m,
36.68 % Pb @ 316 m, and
3773 ppm As @ 302 m
The average concentration of Sb and Bi is 50 ppm.
Some of the anomalous contents of sulphide minerals were at the following depths
(Figures 2.12 and 2.13):
65 vol. % sphalerite @ 315 m,
10 vol. % galena @ 315 m, and
8 vol. % pyrrhotite between 302 and 304 m.
The abundance of the sulphide minerals decreased significantly after 317 m.
Some of the anomalous contents of orange garnet and red garnet are (Figures 2.12
and 2.13):
10 vol. % orange garnet @ 331 m, and
8 vol. % red garnet @ 302 m.
The dominant rock type within the investigated samples of drill hole 4031 is garnet
quartzite and mineralisation shows a regular relationship with it. In Figure 2.13, almost all
textures are equally visible in the sulphide-bearing sections of 4031. The specific gravity
increases within the sulphide-rich zones. |
ADE | Chapter 2-Quantitative Core Logging 63
2.8 Summary
The conventional core logging of the Western Mineralisation and other orebodies
of Broken Hill have focused only on descriptive geological information for different length
of drill cores and they are suffering from lack of quantitative geological information within
the sulphide mineralised samples. In this study, 31 important geological variables
including sulphide and silicate minerals, rock types and textures were quantified for
sulphide mineralised samples and the magnetic susceptibility of 1,928 samples has been
measured. The coordinates of the investigated samples were shown in the collar and the
spatial maps respectively (Figures. 2.1 and 2.2).
The theory of sample volume-variance relationship was used for converting assay
values of HQ and LTK60 cores to NQ cores and different lengths of assayed samples were
recomposited into the equal length of samples in order to integrate the assays data with the
quantitative geological information and the magnetic susceptibility measurements. The
final data base comprises 44 quantitative variables for every investigated sample. Finally,
this chapter compared and contrasted some conventional core logs against their respective
quantitative core logs to demonstrate the resulting quality of information. |
ADE | CHAPTER 3
Quantitative Core-Log Analysis for the Western Mineralisation
3.1 Introduction
In qualitative or descriptive core logging, geologists are limited to conventional or
empirical core logging methods for their geological interpretation. In exploration, the time
lag between qualitative core logging for geology, sample analysis for assay and preparation
of a 3D block model is limited by the progress and quality of mining geology research and
exploration. As mentioned in Chapter 2, conventional core logging cannot fill this gap
because of the many difficulties that exist with its qualitative geological information and
poor capacity to present intricate and detailed relationships between geological features
and geochemical variation. Further, if just a few element concentrations of a mineralisation
(e.g. Pb, Zn and Ag) are integrated for construction of a 3D block model, the final model
will be only useful for mining engineers to estimate tonnage and grade of the orebody and
mineral processing issues. However this may not be of much use for evaluating geological
or geochemical characteristics of a specific type of mineralisation.
The quantitative core-log data set constructed in the previous chapter allows
evaluation of geological information raised from drill cores in a short time and in a
numerical format. The quantitative core log data set can be updated in real time, regardless
of access to assay results and it can be analysed at that stage for geological interpretation.
The quantitative core-log analysis enables the geologist to select relevant core samples for
more detailed geological investigation.
The aim of this chapter is to statistically evaluate geological and geochemical
information derived from samples of drill cores using univariate analysis of descriptive
statistics. The univariate analysis is applied to the quantitative core log data set in order to
describe the data and what is going on the processes of assay values, quantitative
mineralogy and lithology. The univariate analysis involves the examination of a single
variable at a time (e.g. variation of Zn among the assayed samples). The results are
presented in the form of statistical diagrams, comparative bar diagrams and statistical
tables. Moreover, in this chapter, different types of correlation coefficients are used to
measure the degree of relationships between rocks types and Pb+Zn of the Western
Mineralisation and examine the validity and reliability of the estimated galena+sphalerite
with Pb+Zn. |
ADE | Chapter 3-Quantitative Core-Log Analysis 65
The classic statistics of this chapter were calculated by SPSS (SPSS Inc. 2009) and
Minitab (Minitab Inc. 2007) software packages.
3.2 Descriptive statistics
Descriptive statistics are deductive approaches that help us simplify large data sets
of quantitative core logs in a sensible way to recognise drill core samples of interest. This
analysis provides a flexible way to examine geological and geochemical data without
preconceptions. In some circumstances (e.g. when the data set is so large), relying on
geological observation without considering an application of descriptive statistics may
cause biased results, with the loss of important detail or the distortion of the original data.
For this purpose, some univariate analyses of the descriptive statistics were used in this
chapter (Table 3.1).
Descriptive statistics have a strong reliance on graphical visualisation and the
application of descriptive statistics in geology is a powerful tool for:
1. Focusing on significant geological and geochemical indications and comparing and
contrasting them,
2. Providing deeper understanding of the variables,
3. Examining accuracy and precision of the variables, and
4. Summarizing a large data set to simplify information.
However, descriptive statistics cannot alone provide definite answers for geological
and geochemical characteristics; such answers require our final geological judgment. This
chapter aims to concentrate on the application of classic statistical methods rather than
their theoretical mathematics. More details of the theoretical mathematics can be found in
statistical books such as Hoaglin, Mosteller and Tukey (1983) and Tukey (1977). |
ADE | 66 Chapter 3-Quantitative Core-Log Analysis
Table 3.1: A summary of descriptive statistics used in this chapter.
Arithmetic mean (or mean in this study) is often used for measures of central tendency and
indicates that quantitative data tends to cluster around a single value representing the centre
of the data. For skewed data, the arithmetic mean is pulled in the direction of the heavier tail
and in this case is not a true representation of the centre. Therefore, in a skewed distribution,
calculation of the median is preferable.
Median describes the middle of the range of data and separates the higher half of a
population from the lower half. Compared to the mean, the median is less sensitive to
outliers and skewed data and thus often provides a more informative measure of the centre.
A confidence interval for the mean shows the reliability of a value within an estimated range
of values and demonstrates the domain of the variation of the mean value. A confidence
interval is always calculated for a particular percentage confidence level (e.g. 95 %
confidence interval). The ideal confidence interval should be narrower or closer to the
estimated mean value because the distance between a lower and higher confidence interval
indicates the amount of uncertainty for a true mean value. A confidence interval for the
mean is not a robust statistic for skewed data and in these cases, the measurement of
confidence interval for the median is more appropriate.
The COV is a measure of relative variability, equal to the standard deviation divided by the
mean and the result usually is multiplied by 100. As COV is a dimensionless value, it is a
useful tool for comparison of the different dispersion of data with significant different means.
Dispersion in COV refers to the spread of the data around the mean value and it is
calculated by the standard deviation. A greater COV for a variable indicates a higher
variability of the variable relative to its mean.
Skewness is a measure of how much a data set does not have a symmetrical distribution.
Non-skewed distribution or normal distribution has a skewness value of zero. A positive
skewness value (value greater than zero) has a distribution tail that points to the right, which
indicates right skewed data. A negative skewness value has a distribution tail that points to
the left.
Quartiles are values that divide total samples of data into four equal parts that help to
evaluate distribution of a data set and central tendency. The first quartile (Q ) shows 25 % of
1
a data set is less than or equal to the Q value. The second quartile (median) indicates
1
50 % of a data set is less than or equal to the median value and the third quartile (Q ) shows
3
75 % of a data set is less than or equal to the Q value.
3
The amounts of 5 % of sample values that lie at the high end of the total analysed samples
show a higher value than the 95 % probability of the values of the total samples (high
probability or HiP). In fact, the 95 % of total samples have values less than or equal to the
HiP value. In a data set, approximately 5 % of all data sets which contain higher values
relative to HiP value may be suspected to be outlier data. In contrast, the amounts of the 5%
of sample values that are lower than or equal to 5 % value of the total samples (low
probability or LoP) may be considered as background values.
citemhtirA
naideM
lavretni
ecnedifnoC
fo
tneiciffeoC
ssenwekS
selitrauQ
)PiH(
ytilibaborp
%
59
naem
)VOC(
noitairav
)PoL(
ytilibaborp
%
5
& |
ADE | Chapter 3-Quantitative Core-Log Analysis 67
3.3 Bar diagram
In this chapter, several bar diagrams were constructed in order to demonstrate the
variation of the elements (Figures 3.3 to 3.9) and variation of galena and sphalerite (Figure
3.11) within the analysed samples of each drill core in the Western Mineralisation. The bar
diagrams are composed of vertical bars for each drill core that represent minimum and
maximum sample values of an element (or a mineral). Marked on each vertical bar are
mean, median, LoP and HiP values. A significant difference between HiP and LoP values
for an element (or mineral) indicate that it has a large variation between the analysed
samples of the drill core. In a normal distribution, minimum and maximum values of the
vertical bar should be close to LoP and HiP of the vertical bar respectively.
3.4 Statistical results for element concentrations of the Western Mineralisation
The statistical results of the total analysed samples have been given in Table 3.2
and plotted in Figures 3.1 and 3.2. Some statistical points of the table are explained in this
section.
The amount of concentration
The mean concentration of As is more than three times its median concentration. In
Table 3.2, the HiP concentration of As is 22.8 times its Q concentration Table 3.2 also
3 .
shows that Ag has the lowest overall mean and median concentrations. In the Western
Mineralisation, Ag is largely incorporated in an atomic lattice of galena and it also occurs
in tetrahedrite, pyrargyrite, polybasite, argentite, stephanite, dyscrasite, allargentum,
antimonial silver and native silver. According to Table 3.2, the maximum concentrations of
Fe and Bi are more than three times their HiP concentrations. This indicates that 1491
samples are enriched significantly in Fe and 144 samples contain high Bi content.
Skewness
The skewness of elements in Table 3.2 indicates that the 10 elements have positive
skewed distribution. This indicates that the number of samples containing low
concentrations for each element is much higher than those containing high concentrations
for that element. In Table 3.2, As has the highest skewness.
1 Calculated by 0.05 × the number of samples containing Fe concentrations in Table 3.2 |
ADE | 68 Chapter 3-Quantitative Core-Log Analysis
COV
According to the COV results of Table 3.2, As has the highest variability around its
mean concentration and Fe has the lowest variation relative to its mean concentration.
Comparison of the COV values of Zn and Pb in Table 3.2 shows that the variability of Zn
around its mean concentration is 5.7 % 2 more than the variability of Pb around its mean
concentration within their total analysed samples.
Similarities and differences
In Table 3.2, Bi and Sb have some similar values in the number of samples, COV,
median, Q3 and the maximum but they have some differences between their values of
minimum, Q , HiP and skewness. Figures 3.1 and 3.2 show some similarities in
1
distribution of the following elements:
Zn and Pb,
Fe and S,
Bi and Sb, and
Ag, As and Cd.
Figure 3.1 also shows that the dispersion of Cu within the analysed samples is to
some extent similar to Ag, As and Cd in Figure 3.2 but with a wider dispersion.
The statistical results of the analysed samples in surface and underground drill
cores have been given in Tables 3.3 and 3.4 respectively. Some important statistical results
of the Tables are explained in this section.
COV(Zn)COV(Pb)
2 Calculated by 100 from Table 3.2
COV(Pb) |
ADE | Chapter 3-Quantitative Core-Log Analysis 73
The amount of concentration
Tables 3.3 and 3.4 show that the mean and median concentrations of Zn, Pb and Fe
in surface drill cores are smaller than the underground drill cores. In contrast, the surface
drill cores contain higher mean and median concentrations of As, Sb, Bi and S.
Skewness
Comparison of the skewness of Sb and Bi between Table 3.3 and Table 3.4 shows
that Sb has almost similar skewness within both surface and underground drill cores.
However, the skewness of Bi in underground drill cores (22 in Table 3.4) is 5.5 times that
of surface drill cores. In Table 3.4, there are significant differences between maximum
(501.63 ppm) and Q (4.35 ppm) concentrations of Bi within the underground drill cores.
3
COV
The COV results of Tables 3.3 and 3.4 show that the relative variability of Sb and
Bi within the underground samples is 7.8 and 10.6 times those elements within the surface
samples respectively. However, Zn, Fe and Pb show small differences of variation relative
to their mean values between the underground and surface samples.
According to the COV results of Tables 3.3 and 3.4, the variations of Zn and Pb
within the samples of the surface drill cores are respectively 0.8 % and 11.3 % higher than
the samples of underground drill cores. The relative variability of Fe relative to its mean
concentration is almost the same in both of the surface and the underground drill cores.
The COV of Fe in the underground drill cores is also less than the COV results of other
elements.
Similarities and differences
The descriptive statistics of Bi and Sb show significant differences between the
analysed samples of surface (Table 3.3) and underground (Table 3.4) drill cores. Moreover,
Sb and Bi in the surface drill cores have many similar values in their descriptive statistical
results, but in contrast they have many different statistical results within the analysed
samples of underground drill cores. |
ADE | 74 Chapter 3-Quantitative Core-Log Analysis
3.4.1 Interpretation of the bar diagrams of elements for the surface drill cores
Pb, Zn and Ag
In Figure 3.3, the trend variation of mean and median concentrations of Pb and Ag
strongly resemble each other in each drill core. This is natural because major Ag has been
concentrated in the atomic structure of galena in the Western Mineralisation.
In comparison with the overall mean and median concentrations of Ag, Zn and Pb
in surface drill cores (Table 3.3),
1. Drill cores 4002, 4031 and 4033 contain higher mean and median concentrations of
Zn. Moreover, Zn shows high variation within the analysed samples of the drill
cores (see Figure 3.3). This is because of significant differences between their
minimum and HiP concentrations. In those drill cores, the HiP concentrations of Zn
are close to the maximum concentrations of Zn. Sphalerite is the major sulphide
mineral within the drill cores. Although Zn occurs in sphalerite, gahnite and
silicates, acid dissolution and ICP-OES analysis only determined Zn in sphalerite,
2. Drill cores 4002 and 4031contain higher mean and median concentrations of Pb,
and
3. Drill cores 4002 4005 and 4031 contain higher mean and median concentrations of
Ag.
S
The trend variation of S (Figure 3.4) is very similar to the trend variation of Zn
(Figure 3.3). This is natural in the Western Mineralisation because the major sulphide
mineral containing S is sphalerite. Sulphur has also been concentrated in the atomic
structure of a variety of sulfosalt minerals (e.g. gudmundite, bournonite, stephanite,
polybasite, pyrargerite, stibnite and tetrahedrite). The concentration of S (like Zn) shows
great variation within the analysed samples of drill cores 4002, 4031 and 4033 because of
significant differences between their minimum and HiP concentrations. The drill cores also
contain higher overall mean and median concentration of S in comparison with those
values in other surface drill cores (Table 3.3). |
ADE | 76 Chapter 3-Quantitative Core-Log Analysis
Fe
In comparison with the overall mean and median concentrations of Fe (Table 3.2) drill
cores 4002, 4003, 4031, 4033 and 4052 (Figure 3.4) contain higher mean and median
concentrations of Fe. Pyrrhotite is a major sulphide mineral containing Fe in the Western
Mineralisation and other minor sulphide and sulfosalt minerals containing Fe are pyrite,
arsenopyrite, löllingite, gudmundite and tetrahedrite. Galena and sphalerite samples of the
Western Mineralisation contain high Fe. Iron in silicates was not analysed.
Cu
Figure 3.4 displays high mean and median concentrations of Cu among drill cores
3230, 4002, 4022, 4031 and 4052. The concentration of Cu also shows high variation within
the analysed samples of drill cores 4001, 4019, 4020, 4030, 4031, 4045 and 4049 because of
their higher HiP concentrations relative to the HiP concentrations of other drill cores. In the
Western Mineralisation, chalcopyrite is a major sulphide mineral containing Cu. Cupper has
also been incorporated in the atomic structure of a variety of sulphide minerals (e.g. galena,
sphalerite and gudmundite) and sulfosalt minerals (e.g. tetrahedrite, tennantite and
bournonite).
Cd
In Figure 3.5, the drill core 4031 contains the highest mean and median concentrations
of Cd followed by drill cores 4022 and 4042. The mean and median concentrations of Cd in
drill core 4031 are almost 5 and 10 times the overall mean and median concentrations of Cd
respectively within the analysed samples of surface drill cores (Table 3.3). In the Western
Mineralisation, Cd has been concentrated in the atomic structure of sphalerite.
As
In Figure 3.5, the median concentration of As in drill cores 4004, 4032 and 4052 is
about 10 times the overall median concentration of As (100 ppm) in Table 3.3. In Figure 3.5,
significant differences between the minimum and HiP concentrations of As within the
analysed samples of drill core 4048 indicate a high variation of As. In the Western
Mineralisation, the major minerals containing rich As are arsenopyrite and löllingite but As
has also been concentrated in the atomic lattice of galena, sphalerite, pyrite and other minor
sulphide minerals. |
ADE | Chapter 3-Quantitative Core-Log Analysis 79
Bi and Sb
In Figure 3.6, most of the drill cores have mean and median concentrations of ~50
ppm for Bi and Sb. However, there are some drill cores, such as 4015, 4033, 4034 and
4036, where the mean and the median concentrations of Bi are significantly greater than
those of Sb.
In comparison with the overall mean and median concentrations of Bi in surface
drill cores (54.66 ppm and 50 ppm respectively in Table 3.3), drill cores 4002, 4006, 4015,
4017, 4033, 4034 and 4036 contain higher mean and median Bi concentrations. In the
Western Mineralisation, Bi has been concentrated significantly in the atomic structure of
galena, sphalerite, pyrrhotite and other minor sulphide minerals
In comparison with the overall mean and median concentrations of Sb in surface
drill cores (52.47 ppm and 50 ppm respectively in Table 3.3), drill cores 4002, 4006 and
4017 contain higher mean and median Sb concentrations. In the Western Mineralisation
gudmundite is the major mineral containing Sb. However, Sb occurred either in the form of
minor sulfosalt minerals (e.g. tetrahedrite, pyrargyrite and bournonite) or within the atomic
structure of galena, sphalerite, pyrrhotite, pyrite and other minor sulphide minerals. |
ADE | Chapter 3-Quantitative Core-Log Analysis 81
3.4.2 Interpretation of the bar diagrams of elements for the underground drill cores
Zn and Pb
Figure 3.7 shows some similar trend variations of mean and median concentrations
between Zn and Pb. Drill cores 4062 and 4064 contain high mean and median
concentrations of Zn and Pb. There are significant differences between minimum and HiP
concentrations of Pb and Zn in drill cores 4063, 4079 and 4080. This indicates high
variations of Pb and Zn within the analysed samples of the underground drill cores.
Fe
Drill core 4067 has the highest HiP and maximum concentrations of Fe among all
existing surface (Figure 3.4) and underground (Figure 3.7) drill cores. Iron has high mean
and median concentrations within the analysed samples of underground drill cores 4063,
4064, 4066, 4067, 4071 and 4080.
S, Ag and As
In Figure 3.8, the trend variation of mean and median concentrations of S is
comparable with those of Zn and Pb (Figure 3.7) in the corresponding underground drill
cores. However, the trend variation of mean and median concentrations of Ag and As in
Figure 3.8 does not vary in the same fashion. Sulphur displays high mean and median
concentrations in drill cores 4062, 4064 and 4080. Silver and As show maximum variation
within the analysed samples of drill cores 4059 and 4071 respectively.
Bi and Sb
The mean and median concentrations of Bi and Sb are very much lower among the
underground drill cores (Figure 3.9) relative to surface drill cores (Figure 3.6) and vary
more distinctly among the underground drill cores relative to the surface drill cores. Drill
cores 4059, 4063 and 4079 reveal high variations of Sb within their analysed samples
because of their significant difference of minimum and HiP concentrations. Drill cores
4068, 4071, 4074 and 4079 shows great variations of Bi within their analysed samples. |
ADE | Chapter 3-Quantitative Core-Log Analysis 85
3.4.3 Discussion of the statistical results and the bar diagrams of elements
In Tables 3.2, 3.3 and 3.4, a higher COV value for an element (e.g. As in Table 3.2)
relative to other elements may indicate that the element has been affected more than other
elements by retrogression, remobilisation and general cooling of the orebody. In the case
of As, it may have even been added during the Olarian Orogeny as a result of metamorphic
fluid reacting with a redox boundary (i.e. the Western Mineralisation; Plimer 2006b). The
overall mean and median concentrations of elements in Tables 3.2 to 3.4 does not provide
the detailed statistical information about the variation of every element within each drill
core. The bar diagrams of element concentrations provide a better opportunity to visualise
some of the descriptive statistics for drill cores and evaluate their differences and
similarities.
Differences between HiP and the minimum concentration of an element within the
analysed samples of a drill core can better evaluate variation of the element in comparison
with the difference between maximum and minimum concentrations of the element. For
example, in Figure 3.3, the analysed samples of drill core 4031 contain a maximum Zn
value of 28 % and a HiP of 27 %; that means 95 % of the total analysed samples contain
less than or equal to 27 % Zn concentration and this is very close to 28 %. However, drill
core 4039 with a maximum of 19.67 % contains a Zn concentration of less than 2 % for
95 % of the analysed samples. A bar diagram of COV can also compare the relative
variation of an element concentration (or volume percentage of mineral) within different
drill cores (Figure 3.13).
In the bar diagrams, drill cores with similar collar locations (fanned drill holes;
Figure 2.1) show very different mean and median concentrations for some elements. For
example, As in drill core 4048 (Figure 3.5) has several times higher mean, median and HiP
concentrations in comparison with those of drill cores 4040 and 4044. This indicates that
the variation of As changes rapidly in a small scale of space around the drill holes and
there is a significant spatial anisotropic that controls variation of As even in very closely
spaced drill holes. Therefore, quantification of the spatial anisotropic for each element
concentration (or volume percentage of mineral) can better reveal the spatial distribution
and structural variation of the element (or the mineral) within the orebody. More details are
provided in Chapters 6 and 7. |
ADE | 86 Chapter 3-Quantitative Core-Log Analysis
The concentrations of Zn, Fe and Pb vary less than other elements relative to their
mean values in the samples of surface and underground drill cores. This is probably
because the elements comprise the major percentage of sulphide minerals in the Western
Mineralisation.
3.5 Statistical results for sulphide and silicate minerals of the Western Mineralisation
The statistical results of sulphide and silicate minerals have been given in Tables
3.5 and 3.6 and plotted in Figure 3.10 and bar diagram of galena and sphalerite in Figure
3.11. Since the total investigated samples of the Western Mineralisation are limited to the
sulphide mineralised area, the number of samples containing silicate minerals in Table 3.6
is a result of sampling bias. Some important results of Tables 3.5 and 3.6 are outlined
below.
Table 3.5: Descriptive statistics for sulphide minerals.
Sulphide
Number
minerals Mean COV Min. Q Median Q Max. Skewness
of samples 1 3
(Vol. %)
Galena 904 3.5 146.8 1 1 2 4 50 5
Sphalerite 1031 7. 127 1 2 4 8 65 3.1
Chalcopyrite 883 1.8 62.5 1 1 1 2 10 1.9
Pyrrhotite 995 3.2 103.3 1 1 2 4 40 4.5
Arsenopyrite 111 1.8 49.4 1 1 2 2 5 1
Pyrite 87 1.8 63.7 1 1 1 2 8 2.5
Table 3.6: Descriptive statistics for silicate minerals.
Sulphide
Number
minerals Mean COV Min. Q Median Q Max. Skewness
of samples 1 3
(Vol. %)
Gahnite 777 6.3 67.6 1 4 5 8 35 2.1
White quartz 1275 8.9 108.3 1 3 5 10 70 3
Green feldspar 263 5.8 93.2 1 2 5 8 30 2.3
Pink garnet 1261 13.6 65.3 1 5 10 20 50 0.6
Red garnet 64 3.7 68.5 1 2 3 5 15 2.3
Orange garnet 101 11.1 76 1 4 10 15 40 1
Hedenbergite 197 10 94.9 1 5 5 10 60 2.6
Rhodonite 56 6.7 73.7 1 3 5 10 20 1.3
Blue quartz 1520 30 94.5 1 5 15 54 100 0.8 |
ADE | Chapter 3-Quantitative Core-Log Analysis 87
The abundance of minerals
According to Tables 3.5, the investigated sulphide minerals in order of abundance
are: sphalerite > pyrrhotite > galena > chalcopyrite > arsenopyrite > pyrite.
According to Tables 3.6, the investigated silicate minerals in order of abundance
are: blue quartz > white quartz > pink garnet > gahnite > green feldspar >
hedenbergite > orange garnet > red garnet > rhodonite.
According to Table 3.5, the number of pyrrhotite-bearing samples is almost 9 and
11 times that of the samples containing pyrite and arsenopyrite respectively. Pyrrhotite will
be discussed in more detail in Chapter 4.
The volume percentage of minerals
According to Table 3.5, the mean and median volume percentages of sphalerite are
two times those of galena. Pyrrhotite has a higher mean and median volume percentage
relative to those of pyrite and chalcopyrite but the median volume percentage of pyrrhotite
is the same as that of arsenopyrite. In Figure 3.11, sphalerite has high mean and median
values in drill cores 3230, 4002, 4031and 4033 and galena shows high mean and median
volume percentages in drill cores 4005, 4033 and 4044.
Skewness
Tables 3.5 and 3.6 show that all minerals have a positive skewed distribution and
galena has maximum skewness followed by pyrrhotite.
COV
Comparison of the COV results in Tables 3.5 shows that the volume percentage of
galena has maximum variability relative to its mean volume percentage. Comparison of the
COV results in Tables 3.6 shows that the volume percentage of gahnite has minimum
variability relative to its mean volume percentage. In Table 3.5, pyrrhotite has a higher
COV value in comparison with chalcopyrite, pyrite and arsenopyrite. This indicates that
the volume percentage of pyrrhotite changes significantly more than the volume
percentages of chalcopyrite, pyrite and arsenopyrite around its mean value. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.