The Study Of Fault Diagnosis For Industrial Boiler Based On Principal Component Analysis

Fault diagnosis of industrial boiler’s operation process is an important part of the safe operation of the industrial boiler, because the industrial boiler system has the characteristics of high complex degree, variables and strong coupling, so it is hard to set up the precise mechanism model for fault detection and diagnosis. This paper first introduced the basic theory of the principal component analysis, then according to the industrial boiler operational characteristics, using fault diagnosis method based on principal component analysis,establish PCA model by the normal condition of historical data, complete reducing dimensionality and decoupling data and then calculate the T2 and SPE statistic for fault monitoring and complete the fault isolation and identification according to the traditional contributing graph and reconstruction contributing graph. It is proved effective that the method for fault diagnosis of industrial boiler by the fault simulation experiment, and have the certain referenced value.Under the MATLAB computing environment, a fault diagnosis algorithm based on principal component analysis (PCA) is completed via matrix computation and functions. The fault diagnosis system can complete real-time data interaction with the boiler distributed control system (DCS) utilizing OLE for process control (OPC) technology and complete the fault monitoring and diagnosis on line. In the industrial field, it is verified that the online monitoring and fault diagnosis system is effective by the simulation of fault data of process variables.However, the fault diagnose method of industrial boiler based on the traditional PCA may appear misstatement. This paper introduces multivariate principal component analysis (MPCA) concept according to the the characteristics of variable working conditions of industrial boiler operation process, and put forward establishing respectively PCA model for fault diagnosis during the different load range, and enhance the adaptability of working condition for the detection of process variables. According to the analysis of the results of the simulation experiment, it is proved that this method reduce the misstatement, and it has good detective effect to the process variables testing of variable condition.