A Kind Of Data Process Preprocessing And Fault Diagnosis Method In Complex Industry

With the rapid development of the automation technology,the computer technology,and the AI technology,modern industrial process system is developing in the direction of high complexity and integration.The fault diagnosis that aimed at protecting the operating of the modern complex industrial system is much more important,but the collected data of the system face some new phenomenon with a large amount of data,high dimension and variety,the unit information contains less valuable information,which bring new challenge for fault diagnosis.Based on this,this paper carries out the research on data pre-processing and online real-time fault diagnosis,the main research contents are as follows:Fault diagnosis method based on information entropy and relative principle component analysis.Firstly,the information entropy and the information gain method is used to build the metric function to measure the importance of the variables in the function,then build the new data preprocessing model based on the metric function and dimensional normalization model,moreover,combine the data preprocessing model with the relative principle component analysis for fault diagnosis.At last,the performance of this approach is examined in two experiment which are TE process and Wine dataset respectively.Fault diagnosis based on random projection and support vector machine.Firstly,the dimension reduction concept is introduced to solve the computational problem due to the complex distribution and high dimension of the processing dataset for fault diagnosis;then take the coverage analysis to different dimensional reduction method,it can be found that compared with the classic dimension reduction method like principle component analysis,it provides better tradeoff between pair-wised distances reservation and computational time;next,this paper propose an efficient dimension reduction technique named random projection combined with the SVM for fault classification,as a result,it can not only reduce the computation complexity,but also increase the performance of the fault diagnosis.At last,a fault diagnosis experiment of bearing fault dataset is implemented and shows good performance of the proposed method.Algorithm based on kalman filter and online sequential extreme learning machine.Firstly,the online sequential extreme learning machine is introduced to meet the demand of online updating model during the operating of the system;then take the theoretical analysis and experimental verification for the performance of parameter updating between the kalman filter and recursive least square;next,this paper propose an improved OS-ELM algorithm is introduced based on kalman filter(KOS-ELM)for online parameter updating,the proposed algorithm can not only learn the system model quickly,but also can use the new coming information to adjust the model parameter to achieve better performance of fault diagnosis.The simulation based on the regression model validate the effectiveness of the new method.