Abstract
Fault diagnoses of rotating machinery (RM) have played an important role in the safety and reliability of contemporary sustainable manufacturing systems. Extracting features from original signal is a fundamental process for conventional fault recognition performance which needs human intervention and expert knowledge. This article introduces a Modified Moth Flame Optimization with Fuzzy Attention Deep Learning Enabled Fault Diagnosis (MMFO-FADLFD) Model for RMs for the identification and classification of faults. It follows empirical mode decomposition (EMD)-based signal decomposition and principal component analysis (PCA)-based feature reduction processes and fuzzy attention based bidirectional long short term memory (FA-BLSTM) model. Further, the MMFO algorithm is applied as a hyperparameter tuning technique for enhanced fault classification outcomes. The experimental validation of the MMFO-FADLFD model is tested using a dataset and the outcomes are examined under varying aspects and it confirms a promising performance of the MMFO-FADLFD model over other recent methods.