An Optimized Maximum Second-Order Cyclostationary Blind Deconvolution and Bidirectional Long Short-Term Memory Network Model for Rolling Bearing Fault Diagnosis

To address the challenge of extracting fault features and accurately identifying bearing fault conditions under strong noisy environments, a rolling bearing failure diagnostic technique is presented that utilizes parameter-optimized maximum second-order cyclostationary blind deconvolution (CYCBD) and bidirectional long short-term memory (BiLSTM) networks. Initially, an adaptive golden jackal optimization (GJO) algorithm is employed to refine important CYCBD parameters. Subsequently, the rolling bearing failure signals are filtered and denoised using the optimized CYCBD, producing a denoised signal. Ultimately, the noise-reduced signal is fed into the BiLSTM model to realize the classification of faults. The experimental findings demonstrate the suggested approach's strong noise reduction performance and high diagnostic accuracy. The optimized CYCBD-BiLSTM improves the accuracy by approximately 9.89% compared with other methods when the signal-to-noise ratio (SNR) reaches -9 dB, and it can be effectively used for diagnosing rolling bearing faults under noisy backgrounds.