A Unified Approach for Compound Gear-Bearing Fault Diagnosis Using Bessel Transform, Artificial Bee Colony-Based Feature Selection and LSTM Networks

PurposeRotating machinery fault diagnosis is getting more attention nowadays as it improves industrial safety. Most fault diagnosis approaches proposed by researchers can diagnose only one fault at a time. However, compound defects tend to occur more frequently because of the close interaction of many components in industrial applications. Hence, a compound fault diagnosis is required to operate the machinery safely over a long time.MethodsIn this study, a unique Bessel kernel-based Time-Frequency Distribution known as the Bessel Transform is proposed as a technique for the fault detection of a compound gear-bearing system. The Bessel Transform is paired with a feature selection technique based on an artificial bee colony algorithm to choose the features that provide accurate information about the problems. Finally, the chosen features are classified using a long-short memory network.ResultsA case study is used to validate the effectiveness of the suggested approach, and a testing efficiency of 96.75% is achieved.ConclusionThe results show that the proposed transform in compound gear-bearing fault identification is adequate compared with the traditional time-frequency transforms in compound gear-bearing identification.