Failure prediction of mine compressors

Abstract

A typical compressed air system on a deep-level mine requires a combination of compressors to supply sufficient energy to critical underground equipment for safe operation. These compressors are required to run continuously to maintain adequate system pressure. Hence, these systems account for approximately 20-40% of a mine's total energy consumption. It is of vital importance to minimize downtime and maximize availability. Models that involve reliability engineering can greatly improve decision-making to minimize unplanned failures. To achieve this, a unique failure prediction method was developed, based on previous work. The foundation of the previous method was based on a combination of the standard Weibull distribution, the conditional Weibull distribution and basic quantile analysis. It was applied to mine dewatering pumps and successfully predicted the coming failures with good accuracy. In this study, the model was improved and adapted for application on mine compressors. The methodology was validated on various compressors using the leave-one-out cross-validation method. It was found that subsequent failures of the compressors were successfully predicted based on iterative training data sets of the proposed methodology. An average prediction accuracy of 92.30% was achieved across all compressors used in the study