Multi-objective optimization of ANN fault diagnosis model for rotating machinery using grey rational analysis in Taguchi method

In this paper, a monitoring and diagnosis system based on neural network model is proposed (gear bearing-combined faults). Data sets are collected from vibration signals, measured by pseudo-electric accelerometer fixed on various measuring points of an experimental setup (gearbox test rig designed and manufactured especially for this work). Time and frequency analyses are then conducted to determine statistical parameters (root mean square (RMS) value, mean square value, variance, kurtosis factor, crest factor, and clearance factor or margin factor) and spectrums (fast Fourier transform). The time domain parameters and the defects of binary codes are used as input and output data, to train and test the neural network model. Moreover, the L27 Taguchi standard orthogonal array and Grey-Taguchi method are used as multi-objective optimization approaches, to find the best neural network model architecture. The results show the applicability and effectiveness, of the proposed system, for monitoring and diagnosis of rotating machinery. They can be extended efficiently to study other faults and mechanisms.