Research On Performance Assessment And Monitoring Technology Of Multivariable Control Systems

For over 60 years, the theoretical research and industrial practice of the automatic control technology have been developed greatly. The application of automatic control theory brought tremendous benefit to the enterprise and society. However, in practice people gradually found that the performance of the control system was not always in good condition. Even if controllers perform well initially, many factors can cause their abrupt or gradual performance deterioration overtime. In this context, the monitoring of control system performance is of great significance for enterprises'development and is of great concern to engineers.One the one hand, the size of modern continuous process industries is large. In a typical industrial process, there are thousands of control loops. On the other hand, advanced control strategies have been widely used in chemicals, petroleum and other industrial processes. But the studies of performance assessment technology of the advanced control systems show hysteresis. The advanced control systems are essentially multivariate control system. So how to efficiently monitor and assess the performance of multivariate control system become the urgent needs of the modern industrial control process.Based on the above analysis, the research of this paper focuses on multivariable control performance assessment and monitoring techniques, mainly including the following areas.1) In order to simplify the calculation of interactor matrix and reduce a priori process knowledge for performance assessment, this paper introduces the knowledge of diagonal interactor matrix and proposes the performance assessment technology based on diagonal interactor matrix.2) This paper studies multivariable control performance assessment technology based on the user-specified benchmark and proposes an improved computational method for user-specified benchmark, which only requires routine operating data and the estimate of input/output delays. For practical purposes, this is a convenient way as it does not require process model information. And the proposed user-specified benchmark can overcome the defect of the methods based on the minimum variance control benchmark.3) The user-specified benchmark together with the historical benchmark is specified as two parallel benchmarks to form a new integrated performance assessment strategy in this paper. The change both in process dynamics and outside disturbance can cause the control performance degradation. The strategy can distinguish these two reasons and then provide information for controller tuning. The control performance can be assessed and monitored completely through different changes in two kinds of index curves.