The Performance Evaluation Of The Industrial Control Process Variance Algorithm Research

The control system played a crucial role in the modernized manufacturing industry, The performance of the control system will decrease over time due to the lack of maintenance, equipment wear and many other reasons in actual industrial process. The poor performance of the control system would have a serious impact on the quality of products, thus affecting the economic efficiency of enterprises. Therefore, the control performance assessment (CPA) has been widly concerned by academics and industry in past two decades. The key problem of CPA is how to assess the performance of the control system by its operating data.This thesis pointed out the lack of CPA used in actual industrial process by reviewed its development and application. The variance based CPA algorithms for single input single output (SISO) system and the multiple input multiple output (MIMO) system were investigated respectively in this thesis. The main work and the research results are as follows:(1) A recursive algorithm is developed to overcome the shortcomings of linear regression algorithm used in the minimum variance (MV) based CPA for SISO system. Simulation example illustrated the validity of the proposed algorithm.(2) Based on the recursive algorithm, a demo software for the MV-based CPA of SISO system is developed by MATLAB/GUI.(3) The conventional linear quadratic Gaussian (LQG) based CPA for MIMO system can not provide user the performance of each output, an improved strategy is developed to overcome the deficiency of the the conventional LQG method. The MIMO system is first decomposed into multiple multi-input-single-output (MISO) subsystems. Then the LQG benchmark is respectively obtained for each subsystem and the corresponding variance performance assessment indicator is yielded. The performance monitoring result of the overall system can finally be derived by weighting to each subsystem’s assessing indicator. Simulation example illustrated the validity of the proposed algorithm.