Control Performance Assessment Method And System Design Based On Variance Index

In modern industrial production,control systems play a vital role in increasing the productivity of enterprises.However,in actual engineering,the performance of the control system is good or bad,even if there is good performance,it will gradually decline over time,then affecting the efficiency of the enterprise.Control performance assessment based on variance index focuses on the use of running data to evaluate the variance performance of control system,through the evaluation can grasp the status of control system,then provides a reference benchmark for circuits with poor performance.On the basis of analyzing the existing control performance assessment technology,the main research contents are as follows:1)Considering the shortcomings of minimum variance performance assessment method in practical evaluation of univariate control systems,and the limitation that the noise signal is required as a known condition when estimating the index using the FCOR algorithm,a strategy is proposed to evaluate the performance of the univariate control system by taking the generalized minimum variance as the index and using the subspace method to estimate the generalized noise sequence at the future time.Firstly,the design method of the univariate generalized minimum variance controller is introduced in detail,then the generalized minimum variance index is defined as the benchmark by the generalized output under this controller,and the calculation formula of the FCOR algorithm applied to the generalized minimum variance indicator is derived.According to the collected process data,the system is constructed and the generalized output signal is constructed,and the generalized noise signal of the future sequence is identified by the data matrix constructed by the generalized output and the process data,and then the generalized noise signal and the generalized output signal are brought into the calculation formula to obtain the evaluation index.Finally,the effectiveness of the proposed method is verified by using a continuous stirring kettle heater as a simulation object.2)Considering the difficulty of finding the minimum variance performance index due to the complexity of estimating the correlation matrix operation in the real-time performance evaluation of multivariate control systems,a multivariate real-time performance evaluation method based on LQG trade-off curve is proposed.Firstly,system process model subspace and perturbation model subspace is identified through the subspace identification algorithm from the process data,then the optimal control rate of LQG objective function is derived by subspace matrixes.Different control rates can be got by changing the weight parameters in the objective function,then drawing LQG trade-off curves through input variance and output variance of different control rates.Based on this trade-off curve,real-time performance indicators will be obtained through the movement of the time window.To identify potential system hazards early by predicting the future moment performance of the advantages and disadvantages from the performance change trend.Finally,a multivariate model is used as a research object to verify the effectiveness of the proposed method.3)Based on the analysis of the traditional control performance evaluation algorithm,combined with the two improved control performance evaluation algorithms mentioned above,a set of WEB-based process control performance evaluation software was developed.The software can process and save the industrial data uploaded by the user,display the data change trend,basic statistics such as average and variance,and provide various variance performance indicators such as the minimum variance,generalized minimum variance,and LQG related to the circuit according to different system types.The software integrates a variety of control performance evaluation related algorithms,including: stationarity analysis,differential algorithm,delay estimation,time series analysis,system identification,etc.,which can comprehensively and multi-angle mining various types of information related to the performance of the control loop from the uploaded data.In addition to the relevant metric numbers,the information returned to the user is also a visual graphical analysis to help engineers quickly understand the variance performance of the current system.Finally,the software platform was tested with the process data of the distillation column of a chemical group,and the practicality and reliability of the platform were verified.