Fault Diagnosis Technique Based On Multivariable Statistical Process

Automatic control systems for modern industrial processes have been expanding to complexity and larger scale. However, the complexity of system will result in unexpected fault which normally takes much time and expense to figure out and eliminate, and it will also bring large economical loss or even cause human injuries and environmental problems. Therefore, reliability and security are needed to avoid large economical loss brought by accidents and even breakdowns of industrial production. Research on integrated process control system, which includes such roles as control, monitoring, diagnosis, proves to have important theoretical and practical value.Multivariate statistical process control is an important branch in the research of fault diagnosis. In this dissertation, important aspects of process fault detection based on multivariate statistical theory are presented and studied systematically. A new method is presented to find the fault origin from the multivariate statistical aspect. The main contributions of this dissertation are described as follows:1. The reported methods of fault diagnosis are summarized. It is pointed out that fault diagnosis based on multivariate statistical process control has the advantage of no needing of accurately modeling and little effect from the strong correlations among the industrial variables. An introduction of the history and developing tendency on fault diagnosis based on multivariate statistical process control is provided. The mathematical basises of statistical fault diagnosis are introduced: Principal Component Analysis and Partial Least Squares, and their extensions: Multi-way Principal Component Analysis, Multi-block Partial Least Squares and Principal Component Recursion.2. The main statistical variables and multivariate control charts are introduced. The design idea of fault monitoring and fault diagnosis systems based on principal component analysis for multivariable control system is analyzed. A model of PTAoxidation process is talcen as an example to show the result of fault diagnosis based on principal component analysis. Then the inefficiency on fault recognition based on multivariate statistical process control is presented.3. To enhance the ability of fault recognition, a new method for searching the fault origin from the multivariate statistical aspect is presented. The fault origin can be efficiently found out by calculating the probability of each industrial variable making the sample abnormal and analyzing the industrial process.Finally the dissertation is concluded with a summary and discussions of the prospective research on open problems.