Long short-term memory based semi-supervised encoder—decoder for early prediction of failures in self-lubricating bearings

Abstract The existing knowledge regarding the interfacial forces, lubrication, and wear of bearings in real-world operation has significantly improved their designs over time, allowing for prolonged service life. As a result, self-lubricating have become viable alternative to traditional bearing industrial machines. However, mechanisms are still inevitable occur progressively bearings, as characterized by loss lubrication film seizure. Therefore, monitoring stages states these components will help impart necessary countermeasures reduce machine maintenance downtime. This article proposes methodology using long short-term memory (LSTM)-based encoder—decoder architecture on force signatures detect abnormal regimes, aiming provide early predictions failure sliding contacts even before they occur. Reciprocating experiments were performed bronze bushing steel shaft journal custom-built transversally oscillating tribometer setup. corresponding each cycle reciprocating motion normal regime used inputs train architecture, so reconstruct any new signal with minimum error. With this semi-supervised training exercise, could be differentiated from regime, reconstruction errors would very high. During validation procedure proposed LSTM-based model, model predicted signals regimes an accuracy 97%. In addition, visualization error across entire signature showed noticeable patterns when temporally decoded actual critical point, making it possible failure.