Parameter Identification Of A Class Of Non-gaussian Linear Systems And Its Application In Performance Assessment

Non-Gaussian phenomena are ubiquitous in actual industrial processes.With the increasing demand for system control,the Gaussian assumption of interference cannot meet the demand.It is found that the traditional parameter identification and performance assessment methods for Gaussian systems are not applicable to systems with non-Gaussian interference.Based on the application of linear system performance assessment under non-Gaussian interference,this paper studies the parameter identification of non-Gaussian systems,analyzes the existing control loop performance assessment methods,and points out the difficulties and problems encountered in performance assessment under non-Gaussian interference.The importance of studying parameter identification algorithms of non-Gaussian systems for performance assessment is emphasized.On this basis,the main work of the paper is as follows:1.We propose to replace the least mean square algorithm with the least absolute deviation algorithm for parameter identification and performance assessment.Firstly,we analyze and summarize the existing least absolute deviation algorithm solving methods,and point out their advantages and disadvantages through experimental comparison.Then it is applied to the performance assessment of non-Gaussian system.2.We combine the minimum error entropy algorithm with the estimation of distribution algorithm and propose a new double error entropy minimization algorithm.The algorithm was successfully applied to the performance assessment of non-Gaussian systems.It lays the foundation for the performance assessment of the system under non-Gaussian interference.A large number of experiments verify that the proposed framework and algorithm can better estimate the distribution of interference while identifying the parameters of non-Gaussian systems.Both the parameter identification and the distribution identification achieved the desired results.