Dissertation > Excellent graduate degree dissertation topics show

The Application of the Improved Rbf-nn Based on Optimization in Motor Control

Author: ZhangZuo
Tutor: KongFeng
School: Guangxi University of Technology
Course: Control Theory and Control Engineering
Keywords: RBF Neural Network Ant Colony Algorithm Chaotic ergodicity Chaos Ant Colony Algorithm RBF neural network PID control Motor control
CLC: TM32
Type: Master's thesis
Year: 2010
Downloads: 242
Quote: 0
Read: Download Dissertation

Abstract

RBF neural network is an essentially depends on the model of mathematical tools, has a strong learning ability and adaptive ability, suitable for motor such uncertainties and highly nonlinear system control. RBF neural network, however, there are some obvious flaws; strongly depends on the one hand its initial parameter set, once the initial parameters given in error, then will not get optimal neural network structure. The other hand, RBF neural network traditional clustering algorithm clustering high quality, traditional clustering algorithms such as K-means algorithm global search speed, but it's just a rough search process, the need to explore new algorithm with which converged to achieve global optimal search results. This article will use the ant colony algorithm and chaos the ergodicity optimization theory and traditional K-means algorithm merging to resolve the value of the strike of the RBF neural network center. Above algorithm optimized neural network is then applied to the motor control to the advanced nature of the test algorithm and feasibility. First, the ant colony algorithm is introduced into the RBF neural network clustering algorithm, the K-means algorithm global search speed and ant colony algorithm is able to avoid local minima characteristics of traditional clustering algorithm of RBF neural network optimization . RBF neural network optimized for a class of non-linear function approximation. The simulation results show that the neural network optimized nonlinear function approximation effect than the conventional RBF neural network, more conducive to the analysis of nonlinear systems. Secondly, the introduction of the chaos the ergodicity optimization theory and its analysis. Improve the traditional logistic mapping to meet the requirements of the RBF neural network-centric value strike. The chaotic ergodicity speed and global search avoid local extremum characteristics, to optimize the RBF neural network clustering algorithm. Subsequently in this article is to explore the improvement of the ant colony algorithm, chaotic disturbance through increased pheromone to the ant colony algorithm, in order to avoid stagnation generation algorithm. The chaotic ergodicity good RBF neural network and chaotic ant colony algorithm optimization of a class of non-linear function approximation. Simulation results show that chaotic ant colony algorithm to optimize the RBF neural network approximation of nonlinear function better, to prove the feasibility and practicality of adding chaotic disturbance in the ant colony algorithm. Finally, the optimized RBF neural network applications to the PID control, Matlab2007 platform designed RBF neural network PID controller, PID controller using design good dynamic performance of the motor simulation test. Simulation results show that the RBF neural network optimized motor control reflects the high control precision, traceability, robustness, and the ability to ensure that the motor control system has good steady state and dynamic performance.

The Application of the Improved Rbf-nn Based on Optimization in Motor Control

Abstract

Related Dissertations