LAUSR

Compound gear-bearing faults that occur in real-time conditions lead components to fail prematurely. Despite their significance in system failure, these compound faults are rarely studied since extracting accurate information from vibration signals is challenging. Therefore, it is necessary to develop reliable denoising, feature selection, and fault classification technique for forecasting compound faults in a rotor system to assure its durability. This research proposes a multi-stage diagnosis method for the identification of compound gear-bearing failures based on complementary ensemble empirical mode decomposition (CEEMD) based denoising, Artificial Bee Colony (ABC) based feature selection, and Artificial Neural Networks (ANN) based classification. The proposed method is validated through a case study, and the integrated method achieves a classification rate of 95.95%. Furthermore, the proposed method is compared with other denoising, feature selection and classification methods. All the other methods are outperformed by the proposed CEEMD-ABC-ANN method.