Effective Maintenance Planning for Improving the Reliability of Underground Mining Equipment—A Case Study

The productivity and effectiveness of any industry largely rely on the proficient use of both personnel and machinery. Equipment malfunctions and unexpected maintenance significantly lead to production losses and unnecessary investments in novel equipment [1], [2]. Efficient maintenance management is a crucial factor in achieving the continually rising production goals in terms of both quality and quantity [3]. A considerable amount of capital can be lost without a robust maintenance strategy, and there is significant potential to reduce this loss through informed maintenance choices [4]. An unanticipated failure can incur repair costs that are substantially higher than those associated with scheduled maintenance or repairs [5], [6]. Even more critical is the production loss linked to major equipment failure. One approach to lessen the effects of such failure is to enhance the reliability of the machinery. In the business realm, ensuring that equipment operates with a high degree of reliability is a vital component [7]. In underground mining operations, machinery such as Load Haul Dumpers (LHDs) plays a key role in the production strategy. In recent years, the underground metal mines in India have their output fallen short of the expected targets. The primary reason for the decline in production rates within the sector is the inadequate availability of machinery [8]. Improved availability facilitates optimal utilization, thus enhancing the production rate of capital-intensive goods [9]. To increase the availability rate of any equipment, it is crucial to reduce its downtime hours [10]. This can be accomplished by thoroughly assessing the efficiency of the machinery.

Assessing reliability is essential for measuring the effectiveness of any system or device [11], [12]. The effectiveness of equipment largely depends on the dependability of its utilization, the surrounding conditions, the quality of maintenance, the operational practices, and the expertise of the operators [13]. These forecasts are beneficial for structuring suitable maintenance and production plans, performing reliability evaluations, and detecting issues in the production system for the risk assessment process [14], [15].

Reliability assessment is one of the primary methods to validate anticipated results that consistently support the intangible factors like customer loyalty and brand reputation [16], [17], [18]. A phase of inactivity usually results in considerable and major losses. Such losses can occur due to specific faulty parts or components; therefore, it is essential to establish a robust strategy for repairing, replacing, and altering configurations associated with those parts or components [19], [20]. The reliability of a repairable system can be enhanced by applying suitable maintenance strategies [21].

Reliability-based analysis of a shovel and dumpers is the theoretical probability distribution such as Weibull distributions (one parameter (1P), two parameters (2P), and three parameters (3P) distributions) to fit the failure data (time between failure (TBF) and time to repair (TTR)) in any surface miTBFne [22], [23]. This analysis in the mining industry assists in pinpointing preventive maintenance (PM) schedules for production machinery. This leads to a decrease in the total maintenance expenses [24]. Employing statistical reliability techniques provides additional understanding of the maintenance attributes of machinery [25]. The present research has been carried out to evaluate failure, determine the main reasons for failure and their occurrences, identify key subassemblies, and create models for predicting sub-system reliability.

The significance of each specific subassembly or subsystem is assessed by examining time between failure (TBF) data [26], [27]. Data on failure for all four LHD machines has been collected over a designated period from the maintenance logs of an industry, to confirm that each LHD remains operational. The collected data sets were analysed for independent and identical distribution (IID) traits to determine the existence or lack of trends and correlations in the information. The data was regarded as IID when there were no signs of trends or relationships among the data points [28]. The data sets were then employed to assess the goodness of fit allocation via Kolmogorov-Smirnov (K-S) analysis [29]. The suitable distribution models are essential for forecasting the dependability of every single component or subcomponent. The procedure for reliability analysis of a crucial subassembly is depicted in a flow chart (Figure 1) as shown [30]:

The current study has been conducted in one of the leading production industries known as Hindustan Zinc Limited, which significantly contributes to the gross domestic product (GDP) of the Indian economy. A diverse range of large-scale equipment is employed in the industry to extract ore and minerals from underground sources. In addition to Blast Hole Drills, the industry utilizes versatile machines such as Bucket Wheel Excavators, Dragline Excavators, Mining trucks, and LHDs for the loading, transportation, and dumping of loose materials in underground environments. This analysis focused on the performance of Sandvik LHD machines, specifically the LH21, LH22, LH24, and LH25 models (Sandvik LH517 model). Generally, the specifications differ from one model to another, with significant variations in capacities ranging from 1 to 25 tonnes and bucket sizes from 0.8 to 10.0 m$^3$. For example, the Sandvik LH517 has a tramming capacity of 17,200 kg and a bucket size of 7.0 m$^3$, whereas the LH410 provides a tramming capacity of 10,000 kg with a standard bucket of 4.0 m$^3$. Other models, such as the Sandvik LH208L, featuring a capacity of 7.7 tonnes and a bucket height of 3.3 m, are tailored for low-profile operations [25]. Serving as the primary method for transporting ore from mined areas to the main crushing site, LHDs are regarded as mid-level mechanized systems utilized in underground mining operations. Figure 2 is a photo of a LHD machine.

Before performing the intended analysis of the data sets, it is crucial to classify each item of the equipment into a defined number of subassemblies [16]. These categories together with their modes of failure were created, based on the reasoning documented in the maintenance logs kept by the maintenance team. This analysis categorized the selected Sandvik LH517 model LHD machines into seven different subsystems (Table 1).

The information about the historical failure of various LHDs in Table 2 was collected throughout one fiscal year (i.e., from April 2016 to March 2017). The gathered information is displayed as spreadsheets featuring maintenance records compiled by the maintenance team, along with the digital versions of the maintenance data saved on a computer. This data includes the count of failures/failure rate (FF), the TBF and the TTR. Data collected from field visits is shown in Table 3, and the failure rate of each subsystem is represented in Figure 3.

As regards post-data collection from field visits, evaluating the performance of equipment is essential. This evaluation can be performed by computing the KPI, such as availability percentage (AP) and utilization percentage (UP). The availability percentage refers to the duration that the machine can operate to execute its assigned task at its working face. The AP of a machine can likewise be defined as the proportion of available machine hours to the planned working hours. The management of a mine is exclusively accountable for deciding the planned working hours. These hours are considered the complete allotted hours for a certain timeframe of equipment usage. If there are extra hours of work outside the assigned shift, these can be part of the planned working hours. Any unproductive time lasting 15 minutes or less can be ignored. The AP and UP are computed from Eqs. (1) and (2).

UP is defined as the proportion of actual machine operating hours or used hours to the planned working hours. The number of UP changes in accordance with the value of the denominator. Based on observations, the available machine hours are regularly fewer than the scheduled shift hours. Table 4 shows the computed values of AP and UP whereas Figure 4 demonstrates their differences.

The trend test serves to identify the patterns of failure trends within an entire machine or a specific sub-system. This test entails plotting the cumulative number of failures against the cumulative time elapsed between failures. In the current study, the trend test was conducted graphically to ascertain whether the data exhibited a trend, or the failure rate for each sub-system was on the rise, decline, or remained stable. The configuration of the trend plot will indicate whether a piece of equipment is undergoing a reduction in failure rate (improvement) or an escalation in failure rate (deterioration). A non-linear plot signifies the presence of a discernible trend. An upward trend line, characterized by a consistently increasing slope, represents a rise in the failure rate, while a downward trend line, marked by a consistently decreasing slope, signifies a decline in the failure rate [26]. The existence of a trend suggests a correlation. Besides, a serial correlation test was performed to examine the relationship between two variables. The scatter plots depicting the two variables (ith TBF and (i – 1)th TBF) illustrate the correlation between them.

These assessments are conducted to analyze the breakdown of each individual sub-system in order to ascertain the presence of characteristics related to independent and identical distribution (IID). The trend reflected by the gathered data can be evaluated by constructing a graph to illustrate the cumulative frequencies of failures (CTFF) in relation to the cumulative time between failures (CTBF). In conducting a trend analysis, if the data points are arranged in a linear format, it signifies that the examined data sets do not exhibit any trend [26]. In addition, a serial correlation test was executed to investigate the relationship between the ith value of TBF and the (i – 1)th value of TBF. Scatter plots representing the data sets for the ith value of TBF and the (i – 1)th value of TBF reveal the correlation between these two values [30]. The graphs from Figure 5, Figure 6, Figure 7, and Figure 8 demonstrated that the data points were interconnected in a linear fashion throughout the trend examination in the analysis. Consequently, no discernible pattern was detected in the collected data sets. Regarding the serial correlation test, the points appeared to be randomly distributed, indicating the absence of correlation. The results of these assessments implied that the data sets for all subsystems were determined to be trendless, and the points were evenly distributed. Therefore, the IID assumption for the data sets was upheld for each subsystem.

After validating the IID assumption, it is essential to carry out probability distribution of the data sets. This procedure is clarified by examining the best fit/goodness of fit distribution. This study performed the analysis utilizing the ‘Isograph Reliability Workbench 13.0’ software. During the analysis, a variety of parameters were assessed, such as exponential, 1-parameter Weibull, 2-parameter Weibull, and 3-parameter Weibull. Of these, the goodness of fit was found to be most accurately represented by the 2-parameter Weibull and 3-parameter Weibull distributions for different subassemblies (see Table 5). The assessment of the goodness of fit for the data sets was conducted using the Kolmogorov-Smirnov (K-S) test. A reduced significance level ($\varepsilon$) in the K-S test signifies a superior fit. The parametric estimation of theoretical probability distributions was conducted utilizing the Maximum Likelihood Estimate (MLE) technique. Table 6 provides the details of the parameters of Scale $\eta$, Shape $\beta$ and Location $\gamma$ for the optimal distribution function using MLE.

Owing to the trend lessness observed in the data sets, a renewal process has been employed to evaluate the reliability of each separate sub-system. Utilizing the optimal model, the software generated outcomes for the reliability percentage (R) (see Figure 9, Figure 10, Figure 11, and Figure 12 and the unreliability percentage (F), recorded in Table 7. Furthermore, the average time between failures (MTBF) and the average time to repair (MTTR) were calculated to evaluate the metrics of maintainability (Eq. (3)) and failure rate (Eq. (4)), as displayed in Table 8. The MTBF for each LHD machine can be calculated by dividing CTBF by the overall number of failures. Likewise, MTTR was determined by dividing CTTR by the overall number of failures. These computations rely on the most suitable distributions. In this examination, the data sets followed Weibull distributions. The maintainability equation for Weibull distribution is expressed as follows:

Preventive maintenance (PM) involves measures implemented to keep components in a designated state via efficient assessment and prompt failure identification [17]. In this research, PM time schedules were computed according to the expected reliability levels ( Table 9). The determined PM time schedules (Eq. (5)) showed that with a 90% reliability requirement for LH21, maintenance needs occurred every 108.6 hours. Similarly, for LH22, LH24, and LH25, the maintenance periods were 127.14 hours, 138.25 hours and 168.56 hours respectively.

Preventive maintenance is an ongoing process that must be addressed with priority. Regularly servicing equipment through tune-ups can extend its operational lifespan. Moreover, such maintenance practices can avert costly emergency repairs, reduce operational interruptions, and lead to significant long-term savings despite upfront costs involved. It enables the early identification of minor concerns that can be addressed affordably before they develop into serious and costly breakdowns. The strategic implementation of inventory control measures guarantees the availability of suitable spare parts for routine maintenance and common failures, thus reducing equipment downtime during proactive maintenance and repairs. This analysis concluded that preventive maintenance resulted in:

• A decrease in energy and maintenance expenditures of up to 30%;

• A reduction in machinery failures ranging from 35% to 45%; and

• A decrease in downtime of as much as 75%.

The data sets acquired from the four LHD machines were analyzed to evaluate their reliability. The reliability calculations demonstrated that the reliability figures for the Breaking sub-system (SSBr) at 9.46%, the Tyre sub-system (SSTy) at 22.97%, and the Mechanical sub-system (SSM) at 36.49% were notably low. It was determined that SSBr, SSTy, and SSM are the most critical sub-systems in terms of reliability. Insufficient and ineffective maintenance, along with operational practices, lead to unforeseen failures. These problems negatively impact the overall dependability of each machine or piece of equipment.

Forecasting reliability-based preventive maintenance (PM) time intervals will yield vital insights for executing scheduled maintenance. Timely completion of planned maintenance is indispensable for achieving the anticipated lifespan of a machine. When the desired reliability is set at 90%, PM may be performed every 108.62 hours for LH21, 127.14 hours for LH22, and 138.25 hours for LH24. This study observed that various operational and environmental factors necessitate distinct maintenance strategies for different LHD machines. To facilitate effective maintenance planning and organization, it is imperative to evaluate the reliability of each machine or equipment individually.