Adaptive discrete variable structure control using neurogenetic algorithm application to a DC permanent magnet motor

摘要

This paper describes the design and simulation of adaptive discrete variable structure control using Neuro-Genetic Algorithm (NGA). Unlike the conventional Variable Structure Control (VSC) scheme, the input of the system is optimized and calculated every sampling interval with system parameter variations using conventional and NGA optimization methods. The role of Genetic Algorithm (GA) driven by Artificial Neural Network (ANN) is used to optimize the controlled system input and save computational effort. In this work, ANN is used as a smoothing function that learns the relationships between the cost function (fitness function) and the controlled system input in order to interpolate between the trained values. The computer simulation results show that the proposed adaptive controller minimizes and eliminates the chattering in the system input as compared to traditional dither controller based on the variable structure theory. Regulation and tracking problems have been demonstrated using the proposed controller. The application of NGA with VSC theory to a DC Permanent Magnet (DCPM) motor is presented and its robustness to parameter variation is tested.