Inverse Model Control Algorithm Based On Rough Granular Computing

The inverse model control, one of the main method of nonlinear control system research, iswidely used in the practical engineering application. The internal model control(IMC), with stronganti-interferences performance, robustness and traceability, is a practical inverse model controlmethodology for nonlinear system. The modeling of the controlled object model and inverse modelis still confronted with many difficulties for the IMC of complex nonlinear system. Granularcomputing is a new type of intelligent information processing technology. It hews out a new wayfor the modeling of the controlled object model and inverse model. The inverse model controlalgorithm based on rough granular computing is deeply researched as follows:1. The neighborhood rough set and neural network were combined together to be applied to themodeling of the controlled object model and inverse model, the controlled object model and inversemodel based on the rough granular neural network is derived.The problem of information loss in theclassical rough set theory can be solved by the neighborhood rough set theory to some extent. Thedecision tables needed for the modeling is established by the observation data of controlled system.The granular computing based on the neighborhood rough sets theory is used in extracting minimalrule set from the decision tables. Then, the structure and initial parameters are determine by thedecision rules. The controlled object model and inverse model designed by the rough granularneural networks are put into the IMC for on-line control. A case for a second-order differentialsystem is analyzed. Simulation results shows that the IMC based on the granular neural networkshas strong anti-interferences performance, robustness and traceability.2. In the process of rules extraction by the neighborhood rough set theory, each attribute mustbe clustered. The process is complex while the clustering results is important for the rules extraction.Combined the fuzzy rough set and neural network, the fuzzy rough granular neural network isproposed. The granular neural network realizes the identification of the forward and inverse models.The decision tables are also needed for the modeling is established by the observation data ofcontrolled system. The neighborhood rough set is used to clustering of the decision attribute. Then,the initial weight values of granular neural network are obtained by fuzzy rough granular computing.Numerical results demonstrate the accuracy and efficiency of proposed IMC based on the fuzzyrough granular computing.