Research On Sensor Fault Prediction Method Of Electromechanical Equipment Based On Data-driven

The sensors are one of the most important devices in the mechanical and electrical equipments.When sensors fail to operate,they will affect the normal operation of the mechanical and electrical equipments,and even cause great loss to the entire production system.With the support of big data era and the rapid development of intelligent manufacturing,the degree of integration and automation of modern mechanical and electrical equipment also become high.Although the sensor fault detection is becoming more and more difficult to rely on accurate physical model,multivariate statistical analysis methods based on data driven technology can not construct a precise physical model of complex electromechanical equipment of the sensor,only need to analyze measurement data collected by the sensors to complete fault predition.Therefore,it is very important to carry out the research on the sensor failure prediction of electromechanical equipment based on data-driven method.Firstly,this paper introduces the theory and the analysis of sensor fault fault prediction of mechanical and electrical equipment including the classification of fault prediction method and elaboration,analysis of the principal component analysis(PCA),kernel principal component analysis(KPCA)and dynamic principal component analysis(DPCA)method,the basic process of electromechanical equipment for sensor fault prediction of the description and analysis of six kinds of the sensor: deviation fault,drift fault,impact fault,periodic type fault,open circuit fault and complete failure fault.Secondly,research on the draw-wire displacement sensor for the roll position automatic adjustment system in the one time forming of cold-formed steel unit,and the typical fault of sensor deviation fault,drift fault,impact fault and precision decline fault is studied.It proves the validity of the method based on principal component analysis and extension methods of fault prediction simulation and finds that the existing problems,the research of sensor fault prediction on this platform is beneficial to the application of data driven technology to sensor fault prediction.Finally,aiming at the problems of the PCA method,KPCA method and DPCA method in the fault detection,we propose a optimized dynamic kernel principal component analysis(Optimized DKPCA)method,which considers the nonlinear and dynamic characteristics of sensor data feedback and removes irrelevant variables or smaller correlation with principal component contribution rate at the same time and amplitude the limits of variables,which makes the data redundancy greatly reduced.The experimental results show that the optimized dynamic kernel principal component analysis method has lower false negative rate and false positive rate,and the fault detection is more timely and reliable than the traditional data-driven method,this can meet the practical application needs.