Research On Multi-sensor Information Fusion Fault Diagnosis Method Based On AR-SVPMCD

Along with the development of the industrialization, machinery equipment becomes chaining, product quality performance is increasing, once one link fails, all chain fail, which leading to a great loss. Therefore, there’s an important engineering practical significance on mechanical equipment fault monitoring. Based on complicated and nonlinear characteristics of fault monitoring data, for the disadvantages of redundancy of prediction model, inaccuracy of signal feature description and insufficiency of recognition stability, Stepwise variable predictive model based class discriminate (hereinafter referred to as SVPMCD), AR model of time-series analysis, information fusion technology are introduced in the field of fault diagnosis for mechanical equipment condition monitoring respectively, in this paper, multi-sensor information fusion fault diagnosis method based on AR-SVPMCD is studyed, relevant work is as follows:Firstly, because more characteristics leads to a redundancy of variable prediction model, a inaccuracy of signal feature description, then a reducing of recognition rate, a stepwise variable predictive model based class discriminate which based on stepwise regression analysis is proposed. Through the "optimal" variable prediction model, the identification accuracy is improved. The simulation data and mechanical experiments demonstrate the effectiveness of the improved method.Secondly, through the advantages of dynamic modeling and forecasting for short sequences of analysis time series, combining AR model and SVPMCD, AR and stepwise variable predictive model based class discriminate is proposed (hereinafter referred to as AR-SVPMCD). By using the autoregressive parameters of AR modeling as characteristic to train stepwise variable predictive model, the recognition accuracy is greatly improved. The simulation data and mechanical experiments demonstrate the superiority of this method.At last, for the defect of incomprehensive and uncertain of single sensor information, Multi-sensor Information Fusion Fault Diagnosis Method based on AR-SVPMCD is proposed. By getting the preliminary diagnosis by AR-SVPMCD method and the decision fusion diagnosis by D-S evidence theory, the stability of recognition is improved. The mechanical experiments demonstrate the effectiveness of this method.