(362y) Failure Stage Classification and Its Application to Predicting Remaining Useful Life of Bearings

With the improvement in the computational tools, using machine learning techniques in the industry is growing rapidly; and two important applications of it, are in the control of the operation and maintenance of the equipment in a chemical plant, in order to prevent unscheduled downtime. Rotary equipment and those which work at high speed and high pressure usually are the ones that need to be inspected frequently. Among all the rotary equipment, reciprocating compressors usually cost several times more to maintain, and their harsh operating condition leads to frequent failure. One of the most critical parts of reciprocating compressors are bearings, which their failure results in unscheduled downtime in the plant, causing enormous economic loss. The prognostics and health management (PHM) of bearings are crucial to reducing production losses and avoiding machine damage by estimating the failure stage of a bearing. Accurate classification of the failure stage of a bearing can help with predicting remaining useful life of bearings and finding the optimum time for performance inspection and maintenance plans. In this study, using statistical machine learning techniques and feature engineering, we analyze vibration signal data based on an operation condition-corrected health indicator to classify different types of bearing faults with a high degree of accuracy.