Research And Implementation Of Fault Diagnosis And Prediction System For Nuclear Power Equipment

With the overall upgrading of social industrial production,medical transportation and national defense and military fields.Equipment monitoring and fault analysis technology are widely used.Because of the complex background of industrial production,it is very difficult to diagnose and predict faults by accurately establishing mathematical models.The data-driven method,which diagnoses the fault types and predicts the performance trend of equipment only based on the historical data of equipment operation,has attracted extensive attention in the industry.Data-driven methods are also widely used in various fields of industrial production.For the development of operation data analysis and management system of nuclear power equipment onboard,the paper studied the fault diagnosis and prediction of nuclear power equipment based on the Tennessee-Eastman(TE)industrial data for the consideration of data confidentiality.Because of the characteristics of large volume of data,high dimensionality and strong correlation of operation data of nuclear power equipment,principal component analysis(PCA)is utilized to extract the feature information of data,reduce the dimensionality of data,simplify the calculation complexity and improve the accuracy of analysis when analyzing equipment performance based on data characteristics.In fault diagnosis,Fisher discrimination analysis(FDA)and kernel Fisher discrimination analysis(KFDA)are compared to study the problem of multi-classification of faults,and genetic algorithm is used to optimize kernel Fisher discrimination(GA-KFDA)to improve the accuracy of diagnosis.In fault prediction,support vector regression(SVR)is utilized to establish a regression model for time series data of fault to realize fault prediction,and genetic algorithm is utilized to optimize the regression model parameters(GA-SVR),which verifies the good performance of GA-SVR model for data prediction.Finally,PCA feature extraction method,GA-KFDA algorithm and GA-SVR algorithm are integrated into the operation data analysis and management system,and the functions of data management,fault diagnosis and fault prediction of nuclear power equipment is realized.