Research On The Control Loop Performance Assessment Method Under Non-gaussian Disturbances

Nowadays there are a large number of control loops in industrial production and it is of great importance to monitor the performance of these control loops.In case of poor control performance,the problem should be detected and the controller parameters and the system structure should be adjusted in time to ensure normal production.Most of the current performance assessment methods for linear time-invariant systems are based on the assumption that perturbations obey Gaussian distribution,however,nonGaussian perturbations exist in large numbers in real industrial processes.This paper analyzes the existing research results to give a performance assessment method for non-Gaussian systems.The main contents are as follows:1.The performance assessment method of the non-Gaussian generalized cascade control system based on the minimum entropy is presented from the generalized feedback invariant,and its results are compared with those of other assessment methods to analyze the superiority of this method.After that,an iterative identification algorithm based on the controlled autoregressive moving average(CARMA)model is proposed for the non-Gaussian cascade system to obtain the parameters of the system,and the accuracy of this algorithm is verified by simulation.2.A performance assessment method for non-Gaussian multiple-input multiple-output(MIMO)systems is proposed from a data-driven perspective.First,the independent latent structures are extracted from the non-Gaussian MIMO system,and then the virtual independent structures are analyzed and assessed to obtain the control effect of the actual system after the separation of the latent structures.Then,the parameters of the virtual independent process are obtained using the identification algorithm proposed above and the actual controller’s tuning direction is given according to the extraction relation.3.Extending the existing linear quadratic Gaussian(LQG)performance assessment idea to give a performance assessment method based on information entropy to constrain the control inputs and outputs of the system.This method applies to the performance assessment of the system whose control is affected by a special perturbation with an explicit correspondence between variance and entropy.