An Improved Simultaneous Fault Diagnosis Method based on Cohesion Evaluation and BP-MLL for Rotating Machinery

With the requirements for safety and stability of rotating machinery, its fault diagnosis is significantly important. To diagnose simultaneous faults of gearbox and bearing in rotating machinery under different working conditions, an improved algorithm based on cohesion-based feature selection and improved back-propagation multi-label learning (BP-MLL) is proposed in this paper. Cohesion evaluation technique is applied to construct a low-dimensional feature vector by selecting high sensitivity parameters in a high-dimensional vector from time and frequency domain. Improved BP-MLL neural network algorithm considers correlation between labels and adopts ReLU as activation function. To show the effectiveness of the proposed method, hardware experiments are conducted on wind turbine drivetrain diagnostics simulator (WTDDS) for simultaneous fault diagnosis. The experiment reveals that the proposed method can achieve better results than conventional methods under six performance evaluation metrics.