Research On Fault Diagnosis Method Of Boiler System Based On Multivariate Statistical Analysis

Fault detection and diagnosis of heating boiler operation process is an important part of its safe operation.Due to its high complexity,multivariable and strong coupling,it is difficult to establish an accurate mechanism model for fault detection and diagnosis.With the rapid development of data measurement and storage technology,industrial process has accumulated a wealth of process data,which contains rich process information.Therefore,data-driven process monitoring methods have received extensive attention.Among them,the multivariate statistical analysis method has been favored for its unique advantages in dealing with highly coupled and multivariate data,and has become one of the research hotspots.According to the linear correlation and nonlinear correlation between variables in the operation of the boiler system,the fault detection and diagnosis algorithms are studied.Firstly,the fault detection and diagnosis theory based on principal component analysis is studied progressively,including principal component analysis modeling,fault detection and traditional principal component analysis fault diagnosis methods.Then,it is applied to the fault detection and diagnosis of the boiler system operation process,and the simulation experiment is carried out.The results show that although the principal component analysis method has certain accuracy for the fault diagnosis of the boiler system,there are still problems of false alarms or failure to detect the faults,which means that it has shortcomings in fault diagnosis of nonlinear systems.The boiler system is composed of several devices.Different devices operate differently and interact with each other.The method of principal component analysis can not achieve good fault diagnosis.Therefore,a hierarchical modeling and detection method based on hybrid variable correlation analysis is proposed.Firstly,a variable grouping algorithm based on linear evaluation is proposed to decompose the process into linear and nonlinear subsets with different variable correlations.Then,a hierarchical modeling and monitoring fault detection system is proposed.The system not only analyzes the local linear characteristics of the process,but also considers the global nonlinear characteristics of the process,and achieves refined modeling and monitoring.Applying it to the fault monitoring of the boiler system operation process,through simulation analysis,it proves that this method can perform more effective and accurate fault detection on the boiler system than the PCA and KPCA-based fault detection methods.