YE Yongwei, YE Lianqiang, QIAN Zhiqin, REN Shedong

 

(College of Mechanical Engineering, Zhejiang University of Technology, Hangzhou 310014, China)

 

 

Abstract: Aiming at characteristics of automobile coating equipments parameter combinations complicated, strongly coupled, which cause the difficulty in discovering and eliminating the system faults in time, the structure of equipments and fault mechanism were researched, and a new method of fault diagnosis based on PCABP was proposed. The experiment was made on the oven heating system. Firstly, the PCA model of heating system was established, and the optimal signal feature through reducing the dimension of the initial monitoring parameters was extracted. Secondly, sample parameters were taken into training adequately, and the results of the equipment operating status were gotten to complete the fault diagnosis. The experimental result shows that the method of PCABP can reduce the dimension of the initial monitoring parameters without losing the main data, and it also can avoid the defect of the original BP algorithm easily falling into the local extremum. Through the experiment of oven heating system, it testifies the effectiveness of the proposed method in diagnosing the equipment faults.

 

Key words: fault diagnosis; principal component analysis; dimension reduction; BP neural networks; heating system