1 Introduction

South African mining is a massive energy consumer, contributing 15% of Eskom’s electrical revenue in 2021 [1]. The ever-increasing mining depth is a crucial contributor to the excessive electrical consumption of deep-level mines [2,3,4]. The mining depth correlates with increasing underground ambient air temperatures [5, 6], forcing mines to introduce cooling systems to maintain legislated underground working conditions [7,8,9].

At shallower depths, mines rely on ventilation systems as the cooling workhorse [10]. However, at more extreme depths, the efficacy of ventilation systems is reduced significantly [10]. As such, mines integrate ventilation systems with chilled water reticulation networks to counteract the high temperatures [10, 11]. Chilled water reticulation systems typically account for 30–35% of a deep-level gold mine’s total energy consumption [12].

This means that chilled water reticulation systems could account for ~ 4.5% of South Africa’s total energy consumption. In 2022, Eskom generated 227.4 TWh of electricity, of which 176.6 TWh of power was from coal-fired power generation [13]. This means that chilled water reticulation systems consumed 10.2 TWh of electricity at a total cost of R6.2 billion [1, 13]. Furthermore, the emissions in ktpa (kilotonnes per annum) from coal-fired power stations for this consumption is estimated at 67 ktpa of NOx, 133 ktpa of SO2, and 17 327 ktpa of CO2 [14].

Chilled water reticulation systems are complex and immense, with several kilometres of pipelines typically found within a deep-level mine [15]. Unfortunately, these chilled water reticulation systems are prone to wastage, often exceeding 30% of the total consumption of these systems [16,17,18]. Due to the complexity and size of these networks, this wastage often goes unnoticed [19, 20].

Several techniques exist for wastage identification in deep-level mining: measured baselining [18, 21,22,23,24,25,26], zero-waste baselining [27,28,29,30,31,32], leak detection [18, 21, 24, 25], and control valves [16, 18, 33]. Measured- and zero-waste baselining aim to use data to generate a reference point [25, 27]. Comparing this reference point to actual water usage provides a platform for wastage identification [25, 27].

Leak detection and control valves are primarily wastage reduction techniques that, once implemented, can be used retroactively to identify wastage (i.e. the flow reduced using these techniques would be classified as wastage) [25].

Some studies have compared or applied multiple of these techniques [18, 19, 24, 25, 27], although not all have been compared in a single study. A comparison of these techniques is required to appropriately define the best use case for, and effectiveness of, each. Thus, a need exists to compare these four techniques using a single comparative reference. To fulfil this need, this paper will.

  • Investigate water wastage identification and reduction techniques

  • Apply them to a deep-level mine

  • Compare the wastage identification effectiveness of these techniques

A case study research design methodology [27, 34] is used to accomplish these objectives. The methodology is selected to provide a platform to demonstrate and compare each method [27, 34]. The case study where the research took place is a deep-level gold mine in South Africa.

2 Wastage Identification Techniques—Overview

2.1 Measured Baselining

Measured baselining is defined by using measured data to develop a reference point [21,22,23,24,25,26,27], which can be used to identify system inefficiencies [25]. In the deep-level mining industry, two forms of measured baselining are commonly used for wastage identification: intensity benchmarking (Section 2.1.1), and historic baselining (Section 2.1.2).

2.1.1 Intensity Benchmarking

Intensity benchmarking seeks to contextualise underground water usage [23, 26]. Intensity is the ratio of water usage per unit production and provides a frame of reference for the water usage within the mine [18, 23]. Figure 1 provides an example of intensity benchmarking on a deep-level mine compressed air system.

Fig. 1
figure 1

Typical intensity benchmarking example [22]

Full size image

Intensity benchmarking is helpful as an initial point of identification of wastage or as a wastage comparison point between different focus areas (as in Fig. 1). The drawback of this method is that it cannot be used to quantify the wastage, nor can it be used to generate a flow (wastage) profile to indicate at what point the wastage is occurring. This is where the second portion of the measured baselining is required.

The typical method for intensity benchmarking is shown in Fig. 2. The process starts with data acquisition (step 1) which aims to collect all required data for the intensity calculations in step 2. As discussed, intensity is the ratio of water usage per unit production defining the data’s scope. The total water usage and ore production data are required for each considered area (at minimum, the reference, and focus areas).

Fig. 2
figure 2

Simplified typical method for intensity benchmarking (adapted from [22, 23])

Full size image

Using the data, step 2 calculates the intensity of each considered area. Each intensity can then be compared to identify the extent of wastage in the area. This process can also be used with historical data to indicate increased or decreased wastage within a considered area [27]. The type of data collected during step 1 is determined by the focus of the intensity benchmarking performed (historic or across areas).

2.1.2 Historic Baselining

Several studies have used measured baselines to establish profiles (flows, pressures, energy) to demonstrate savings [18, 21, 27,28,29,30,31,32], although a measured baseline can also be used to identify wastage [25, 27]. A historic 24-h flow profile can be generated using the measured flow data. This profile can be generated using a Supervisory Control and Data Acquisition (SCADA) system and available flow metres. Alternatively, one could use portable flow metres to measure a point of interest across different time periods to establish a historic flow profile.

Compared to the actual flow profile, this profile can be used to identify and quantify wastage. The benefit of comparing flow profiles is that the time of day for the wastage can also be identified alongside the wastage magnitude [27, 31].

The main drawbacks to historic baselining are twofold. Firstly, using a historic baseline assumes that the water consumers remain consistent, which is not an accurate assumption [25]. Secondly, historic baselines already include wastage. This means that the wastage identification potential of historic data is reduced as any wastage included in the historic baseline is not included in the identified wastage. A limitation with these baselines is that historic profiles cannot be drawn if flow metres are malfunctioning or absent.

The only data required for historic baselining is water usage (as per Fig. 3). Both the historical- and current actual flow data must be acquired and used to establish a 24-h flow profile in step 2 (such as the “current” profile in Fig. 7). The actual and historic flow profiles can then be compared in step 3 to identify wastage.

Fig. 3
figure 3

Simplified typical method for historic baselining (adapted from [25, 27])

Full size image

As with intensity benchmarking, the historic flow profile can differ based on the focus of the wastage identification required (year-on-year wastage versus month-to-month). The comparison differs from that of intensity benchmarking, as the time of the wastage indicates the origin of the wastage, so this is noted from the comparison.

2.2 Zero-Waste Baselining

An alternative baselining method, zero-waste baselining, seeks to determine the minimum consumption for maintaining peak operational output [30, 35]. The application of zero-waste baselining to deep-level mining is similar to historic baselining. A minimum flow consumption profile (zero-waste baseline) is established based on the consumer-specified water requirements [27,28,29,30,31,32]. This is compared to the actual flow profile to quantify wastage, as shown in Fig. 4 [31]. In this figure, the process was applied to a compressed air system, as zero-waste baselines have not yet been applied to deep-level mine water systems and reported in the literature. However, the principles for zero-waste baselining are similar for water- and compressed air systems.

Fig. 4
figure 4

Zero-waste baseline compared to actual flow profile [31]

Full size image

The main benefit of zero-waste over historic baselining is the improved wastage identification potential [27]. Consequently, the main drawbacks are more data requirements and a slower profile generation process.

The method (Fig. 5) for zero-waste baselining is similar to historic baselining. The most significant differences are found in steps 1 and 2. Step 1 shows that zero-waste baselines require more information than the other considered methods. This is because water usage is required to establish a flow profile for comparison to the zero-waste baseline in step 3 (as with historic baselining).

Fig. 5
figure 5

Simplified typical method for zero-waste baselining (adapted from [27,28,29,30,31,32])

Full size image

However, production information, standards, consumer specifications, and schedules are required to establish the zero-waste baseline in step 2 [27,28,29,30,31,32]. This information is used to determine the minimum flow required to operate all consumers, and these minimums are then plotted as a 24-h profile using the consumer specifications (as shown in Fig. 4).

2.3 Leak Detection

The most common wastage reduction technique is leak detection [18]. In deep-level mines, leak detection is primarily performed through visual inspection due to its ease and cost-effectiveness [18, 21]. Leaking pipes are a common issue in deep-level mines and can constitute a significant portion of the present wastage [18].

Once the leak detection procedure is completed, the impact will be noted in the flow measurements. Through this, leak detection can be used as a form of wastage quantification by comparing the new flow profile to one including the wastage [18, 21, 25]. The core drawback to leak detection as a wastage quantification technique is the time required to perform the visual inspection and repair the pipeline.

The methods followed for the previous techniques have been consistent between the focus of each step. This trend changes with leak detection, where the data acquisition step moves to after the techniques focus (as shown in Fig. 6).

Fig. 6
figure 6

Simplified typical method for leak detection (adapted from [18, 21, 24, 25])

Full size image

In step 1, two activities drive the leak detection aspect. Leak audits identify the location and magnitude of leaks (wastage) present, and then repair operations are conducted to eliminate these leaks.

As with historic baselining, leak detection only requires water usage data. This data establishes and allows for comparing the flow profile before (actual) and after the leaks are repaired. Through this comparison, water wastage due to leaks is identified.

Typically, leak water would be a consistent flow throughout the day, forming part of the system’s baseload water demand [27, 33]. As such, identifying increased water usage over the expected baseload would indicate water wastage due to leakages.

2.4 Control Valves

The final wastage identification technique considered in this paper is the use of pressure-based control valves. This method reduces unwanted water flow through pressure reduction [16]. The technique is primarily used in urban water systems for leakage flow reduction [36] but has been adapted to the mining sector for a similar end [16].

Two common locations for control valves are main inlets and the inlets to smaller working places [16, 18, 33]. The latter allows water consumers that are not closed by the end of a shift to be controlled without impacting cooling infrastructure. The former allows for a broader reduction of leakages within a mining system [16, 33]. Figure 7 shows control valves and leak detection applied to a deep-level mine water system.

Fig. 7
figure 7

Leak detection and control valves for wastage reduction (adapted from [18])

Full size image

Figure 7 shows that control valves can reduce waste after thorough leak detection. Another aspect of control valves that Fig. 7 highlights is that they only provide a benefit in the blasting shift (16:00 to 21:00 in Fig. 7) so as to not impede any operational capacity. As such, control valves’ potential wastage identification and reduction potential are limited.

Further, both leak detection and control valves fail to identify the true extent of the wastage in the system (as evidenced by control valves identifying additional wastage over leak detection).

The method followed for wastage identification using control valves is shown in Fig. 8 and is similar to that followed for leak detection. The only notable differences are found in the required activities for step 1. Instead of leak audits and repairs, this technique requires the installation of control valves at key identified locations (as discussed in Section 2.4). The water flow rate is controlled using these valves. Step 3 sees the pre- (actual) and post-control flow profiles compared to identify the wastage eliminated using the valves.

Fig. 8
figure 8

Simplified typical method for control valves (adapted from [16, 18])

Full size image

2.5 Summary of Methods for Wastage Identification Techniques

Figure 9 compares the simplified methods used for each of the wastage identification techniques discussed in Sects. 2.1–2.4. The figure first shows the three techniques that focus on wastage identification, followed by those that identify wastage through wastage reduction activities. The two groups are aligned with the different orders of their respective wastage identification processes.

Fig. 9
figure 9

Summary of wastage identification technique methods

Full size image

3 Wastage Identification Techniques—Application

This section applies the methods discussed in Section 2 to a mutual case study (per the case study research design methodology). This application of the techniques will demonstrate the typical results expected from each and allow for a platform for comparative analysis.

The selected case study is a conventional deep-level gold mine in South Africa. The mining operations are primarily driven using compressed air and the mine has seen a notable increase in water wastage over the past few years, prompting investigation. This mine’s water system is representative of most deep-level gold mining operations’ water systems in South Africa.

3.1 Measured Baselining

3.1.1 Intensity Benchmarking

The tonnes of ore produced, and water usage data are acquired from the mine SCADA system and shown in Fig. 2. The water usage data is logged on in 2-min intervals and was collected over a calendar month, although the data is available for a 5-year rolling period. The case study mine has four levels, so the data must be acquired for each. A monthly water intensity for each level is calculated by dividing the total water usage by the total tonnes produced by each level. These intensities are plotted and compared in Fig. 10.

Fig. 10
figure 10

Case study—water intensity benchmarking

Full size image

From Fig. 10, the intensity of levels A–C is relatively consistent. However, level D has a significantly increased intensity, indicating a higher magnitude of wastage [22, 27]. Unfortunately, the actual wastage periods and quantity cannot be determined from Fig. 10. The remainder of the applications will be on level D for the considered month used for the intensity calculations.

3.1.2 Historic Baselining

Historic baselining should be able to highlight the quantity and time of wastage not shown by the intensity benchmarking. The actual flow profile is the weekday average flow profile for the month considered in Section 3.1.1, while the historic flow profile is for the same month in the prior year. These profiles are plotted and compared in Fig. 11.

Fig. 11
figure 11

Case study—historic baseline compared to actual profile

Full size image

The weekday actual daily average flow is 68 l/s, while the historic baseline has an average flow of 66 l/s. These flows would identify an average wastage of 2 l/s (3%). From Fig. 11, the profiles are closely matched with the bulk of the wastage outside the peak flow periods, and the identified wastage is lower than expected from Fig. 10.

3.2 Zero-Waste Baselining

The zero-waste baseline requires the consumer specifications and operating schedules over the production- and water usage information acquired in Section 3.1. The zero-waste baseline is established and compared to the actual flow profile in Fig. 12.

Fig. 12
figure 12

Case study—zero-waste baseline compared to actual

Full size image

The average flow for the zero-waste baseline is 48 l/s, identifying an average wastage of 20 l/s (30%). This wastage is as expected, based on Fig. 10, highlighting that the drawback to historic baselining is that the historic baseline includes wastage in its profile. Figure 12 shows wastage throughout the day, meaning that both leak detection and control valves would identify and reduce wastage on level D.

3.3 Leak Detection

The maximum leak detection can be estimated using the zero-waste baseline. As discussed in Section 2.3, leakage flows are expected to remain consistent throughout the day. Thus, the difference between the zero-waste baseline and actual flow profiles during the off-peak periods (05:00 to 08:00 and 15:00 to 23:00) can be used to determine the leaks in the system.

From Fig. 12, the lowest difference between the actual and zero-waste baseline in the off-peak periods occurs at 05:00 and is 14 l/s. Applying a consistent 14 l/s (20%) reduction to the actual flow profile yields a theoretical maximum leak detection wastage identification, and the resulting profile would show the impact of the leak detection activities.

3.4 Control Valves

Similar to leak detection, the maximum wastage identification for control valves can be determined using the zero-waste baseline. From Section 2.4, control valves are applied during the blasting shift (15:00 to 23:00 for level D). Comparing the zero-waste baseline and actual flow profiles during this time shows a minimum difference of 17 l/s (at 21:00).

“Controlling” the actual flow profile to that of the zero-waste baseline during the blasting shift generates a flow profile representative of the potential impact of control valves on level D. This profile is compared to the actual profile in Fig. 13.

Fig. 13
figure 13

Case study—actual compared to the valve-controlled flow profile

Full size image

The control valve flow profile from Fig. 13 matches the expectations from Fig. 7. The flow reduction during the blasting shift averages 19.7 l/s (32%) for an average daily wastage of 6 l/s (8%).

4 Discussion

4.1 Case Study Comparison

Figures 10 and 11 show the results of applying the four considered wastage identification techniques to a mutual case study. Table 1 summarises the requirements and wastage identified through each of the applied methods.

Table 1 Summary of method requirements and wastage identification
Full size table

From Table 1, these methods identified wastage ranging from 2 to 21 l/s (3–30%), with historic and zero-waste baselining identifying the least- and most wastage, respectively. Intensity benchmarking is not included in these wastage ranges as it cannot quantify the identified wastage.

In contrast, zero-waste baselining took the longest of the wastage identification techniques (for the data acquisition) but showed the benefit of the additional time spent versus the historic baselining. It should be noted that the proper application of leak detection and control valves requires significantly more time than the theoretical application used in this study [16, 18, 21, 25].

The most reliable/accurate representation of the wastage from the considered techniques should be the zero-waste baseline, as the other techniques neglect elements of the wastage.

Leak detection only quantifies the baseline wastage [27, 33] and neglects the wastage during the drilling shift due to equipment misuse or poor practises. Control valves can only quantify the wastage during off-peak periods [16, 18] which neglects the bulk of the misuse and baseline wastage. Historic baselining does not account for the wastage present at the time of the historic baseline.

Regarding the time required to complete each technique, it is noted that historic baselining is the fastest technique, followed by intensity baselining which requires production data alongside the data required for historic baselining. The second fastest technique is zero-waste baselining which requires specific consumer data. This means that the wastage identification focused techniques are generally faster than the wastage reduction techniques (as expected). This is because these techniques only require data to identify the wastage.

The times for wastage identification through leak detection and control valves are broad estimates as it depends on the number of areas that need to be audited or have valves installed. Furthermore, the mines’ response time is often slow, requiring months to install a single valve or fix a single leak.

The process used to apply these methods to this case study would not be influenced by alternative mining conditions (such as size, depth, or location). These factors would however influence the time required for each technique. For example, a deeper mine with more active levels would require more data acquisition and analysis. A change in mining technique (e.g. conventional to mechanized) could result in a larger impact on the waste identification. However, this is recommended for further study.

The results from each application align with the expectations from the literature (Section 2). Further, the wastage identification of the methods correlates with the increased intensity of level D. A limitation within this case study comparison is that the leak detection and control valves are pseudo-applications using the zero-waste baseline.

This pseudo-application means that the actual wastage identification from each is not represented. Conversely, this should represent the maximum wastage reduction potential, as the zero-waste baseline represents the minimum allowable flow [35]. This application thus shows how the zero-waste baseline can be used as a theoretical analogue for leak detection and control valves. Using the zero-waste baseline in this manner can allow for wastage reduction predictions and pre-empt the expected benefits of applying either leak detection or control valves.

Despite this limitation, the study objectives are met as the wastage identification techniques applicable to deep-level mines have been compared using a mutual case study. This comparison allows for the effectiveness of each method to be determined, which shows zero-waste baselining to be the optimal method for determining the true extent of wastage.

4.2 Proposed Application Method

Figure 14 proposes a method for wastage identification and reduction that combines the considered techniques. This method represents the findings of this study and aims to provide an optimised process of wastage elimination in deep-level mines.

Fig. 14
figure 14

Proposed wastage identification and reduction procedure

Full size image

The method comprises three steps: wastage identification, quantification, and reduction. Each of these steps utilises at least one of the techniques to further the goal of effective and timely wastage elimination. The method starts in step 1 with intensity benchmarking (as described in Section 2.1.1) as a quick wastage identification point. Should the intensity be above the reference (as with level D in Section 3.1.1), the method moves on to step 2: wastage quantification.

The comparative case study shows that zero-waste and historic baselining have the same outcomes (wastage quantification). However, zero-waste baselining better captures the extent of wastage and allows for a more reliable platform for wastage quantification. Thus, the wastage quantification (step 2) utilises zero-waste baselining. Should the necessary data for the zero-waste baseline not be available, historic baselining can be used instead.

Following the zero-waste baselining process (per Section 2.2), step 3 first considers the off-peak wastage. Off-peak wastage is primarily due to leakages [18, 25], and as such, leak detection is initially employed when off-peak wastage is detected. Further, off-peak wastage is considered before peak wastage, as any leakage flows will be detected in the off-peak but will influence the peak wastage (as in Section 3.3). If, after leak detection, there is remaining off-peak wastage, then control valves are used to reduce the wastage further (as in Fig. 7).

Following the control valve application (or if there was no off-peak wastage/no remaining off-peak wastage), the peak wastage is considered. The techniques investigated in this study are focused on off-peak wastage, and as such, if peak wastage remains, this is likely due to poor water usage discipline or water misuse [27]. Either way, peak usage wastage is the final consideration of the proposed method, resulting in the usage discipline requiring investigation or no further wastage reduction.

It is further proposed that the wastage reduction step be theoretically applied using the zero-waste baseline (as in Sects. 3.3 and 3.4). This theoretical application will allow for the maximum potential benefits of the wastage reduction to be quantified and the cost–benefit thereof to be considered before implementation, much as one would use simulations [37]. Figure 15 shows the results following a theoretical implementation of the proposed method to the level D case study.

Fig. 15
figure 15

Case study – wastage identification comparison

Full size image

Figure 15 shows the reduced wastage profile, with an average flow rate of 52 l/s. This profile would indicate a total wastage reduction of 16 l/s (77% of the total wastage identified from the zero-waste baseline) using leak detection and control valves. This reduction would represent approximately R 1.05 million per annum in pumping cost savings [38] for the case study mine. This represents 7.5% of the case study mine’s total annual pumping costs and 0.53% of the mine’s total annual electrical costs.

In this case, the control valves only contributed an average of 2 l/s (12.5%) to the reduction. This minor reduction shows a further benefit to using the proposed method (Fig. 14), as the cost–benefit of wastage reduction can be evaluated before committing the required resources to implement the reduction. For the control valves, the expected savings yield is approximately R 130, 000, which can be compared to the valves’ cost to determine implementation viability.

The proposed method (Fig. 14) should improve the process followed in wastage elimination in deep-level mines, and the application to the case study mine shows a wastage reduction of 77%.

It should be noted that this theoretical application represents the best-case scenario for wastage reduction, as the zero-waste baseline predicts the minimum usage, i.e. the highest opportunity for wastage reduction. In a practical application only a portion of this wastage can be reduced. However, the benefits of the theoretical application can still allow for an evaluation of the maximum benefits before committing to a wastage reduction implementation.

Several potential challenges could arise in a practical implementation of the proposed method. Reliable- and accurate data is required to conduct the wastage identification (step 1) and wastage quantification (step 2) steps. If no data is available, or the data is not credible, then the core of the proposed method is not feasible. Next, the consumer specifications required for the zero-waste baseline can be cumbersome to acquire depending on the information availability on the mining site.

For wastage reduction (step 3), the leak detection procedure has limitations regarding the reliability of information acquired, as leak detection often relies on spot measurements to locate the leaks [25]. This means that the magnitude of the identified leaks can vary based on several underground variables.

Finally, implementing control valves underground requires continuous planning- and monitoring efforts as well as capital outlay for the required infrastructure. This could be viewed as problematic for mine management. However, by first applying the zero-waste baseline theoretically to determine the potential cost benefit, this issue can be overcome.

5 Conclusion

Deep-level mine chilled water systems are energy-intensive and prone to wastage. While several techniques exist to identify this wastage, these have not been compared in a single application. Thus, a need exists for such an application to determine the effectiveness and optimal application for each technique. This study investigates four techniques used for wastage identification in deep-level mines: measured baselining, zero-waste baselining, leak detection, and control valves.

This investigation provides an overview of the techniques and the typical method followed by each, including the requirements and activities expected from each technique. Each technique is applied to a single mutual case study to provide a comparative platform. The application shows wastage ranging between 2 and 21 l/s (3–30%), with intensity benchmarking identifying, but not quantifying, wastage on the case study. The zero-waste baseline theoretically estimates the maximum wastage reduction from leak detection and control valves. Furthermore, historic and zero-waste baselining quantifies the least and most wastage, respectively.

Following the comparison, an optimised method is proposed for wastage elimination using the considered techniques. This method comprises three steps: wastage identification (intensity benchmarking), quantification (zero-waste baselining), and reduction (leak detection and control valves). The method is theoretically applied to the same case study used for the comparative analysis. It reduces the wastage by 16 l/s (77% of the identified wastage), resulting in annual pumping savings of R 1.05 million.

A complete comparative analysis is recommended to be performed where leak detection and control valves will be implemented to obtain a “real-world” wastage identification from these techniques. Furthermore, future studies could apply the proposed wastage identification and reduction method to various case studies to further validate the benefits thereof.

Further studies should investigate an alternative scheduling method for zero-waste baselining. The method followed in this study relied on the planned underground mining schedules, but the actual operations often do not follow these schedules. This could result in inaccurate wastage quantification due to wastage being attributed to mining operations.