Bearing Malfunction Detection Based on Simplified Fuzzy Adaptive Resonance Theory Map Neural Network and Discrete Wavelet Transform

Malfunction detection is a challenge in condition based maintenance (CBM) implementation to minimize uncertainty and cost in maintenance activity. This study proposes a model that combines simplified fuzzy adaptive resonance theory map neural network (SFAM-NN) and discrete wavelet transform (DWT) to conduct CBM for bearing malfunctions. SFAM-NN uses time domain features, such as RMS, kurtosis, RMSEE, and AC Power of vibration signal and AC power of autocorrelation of wavelet coefficient from each sub-band, as inputs. The model provides 22 output categories: one is for healthy condition and the other 21 are for a varying degree of bearing degradation and malfunction type. A smoothing and voting algorithm is proposed to find malfunction classification. Experimental results show that the proposed model can reliably malfunction type.