Gearbox fault detection using entropy-based feature extraction and hybrid classifier

Gearbox fault diagnosis is a crucial aspect of maintenance and reliability in automobile engineering. In automobile vehicles, the gearbox is a vital component that facilitates efficient power transfer from the engine to the wheels, enabling optimal performance, speed control, and fuel efficiency. Therefore, early diagnosis of gearbox faults is crucial to avoid severe damage to the gearbox and other parts. This study aims to develop a novel hybrid deep learning method for gearbox fault detection and classification. It acquires statistical characteristics, higher-order statistical features, modified log-energy entropy, modified Renyi entropy, and the Shannon feature. A bidirectional long-short-term memory (Bi-LSTM) and a recurrent neural network (RNN) detect faults. Opposition learning combined with an artificial humming-based crow search algorithm (OAHCSA) determines the RNN weights. The outputs are further averaged to obtain more accurate findings. The proposed OAHCSA with the HC model achieves 99.62% accuracy at 15 Hz, 98.85% at 20 Hz, 99.23% at 25 Hz, and 99.23% at 30 Hz. The findings show that the proposed method has the potential to be an effective alternative for precise gear fault detection.