Feature Ranking and Differential Evolution for Feature Selection in Brushless DC Motor Fault Diagnosis

A fault diagnosis system with the ability to recognize many different faults obviously has a certain complexity. Therefore, improving performance of similar systems attracted much research interest. This article proposes feature ranking and differential evolution for selection in BLDC diagnosis. First, this study used Hilbert–Huang transform (HHT) extract features four types brushless DC motor Hall signal. When there is fault, symmetry signal will be influenced. Second, we based on distance discriminant (FSDD) calculate factors which base category separability select have positive correlation types. The were entered sequentially into two supervised classifiers: backpropagation neural network (BPNN) linear analysis (LDA), identification results then evaluated. input classifier was derived from FSDD, optimized rank using (DE). Finally, verified motor’s operating environment simulation same by adding appropriate signal-to-noise ratio magnitudes. obtained an accuracy rate 96% when 14 features. Additionally, experimental show that proposed robust anti-noise ability, 92.04%, even 20 dB white Gaussian noise added Moreover, compared established discrete wavelet (DWT) variety classifiers, our higher fewer