补偿模糊神经网络自适应逆控制的仿真研究

Simulation Study of Adaptive Inverse Control with Compensation Fuzzy Neural Network

该文应用的补偿模糊神经网络(CFNN)是结合补偿模糊逻辑和神经网络的混合系统。由于引入补偿神经元使网络容错性更高,系统更稳定;同时模糊运算采用动态的、全局优化运算,并在神经网络学习算法中动态优化补偿模糊运算,使网络更适应,训练速度更快。将补偿模糊神经网络与自适应逆控制原理结合应用到某位置伺服系统噪声消除控制中,并同用BP网络,传统PID控制和常规模糊神经网络控制效果比较来证明此方法的优越性。仿真结果表明补偿模糊神经网络自适应逆控制在缩短训练时间,提高控制精度等方面都有显著改善。

The compensative fuzzy neural network （ CFNN）, which combines compensative fuzzy logic with neural network is introduced. Compensation neurons make the system much more stable. Meanwhile fuzzy computation is dynamic and global optimized, and compensative fuzzy computation is optimized dynamically in the study algorithm of neural network. Therefore the network is much more adaptive and speed is much faster. CFNN is combined with inverse control principle to cancel disturbance of some servo system. BP neural network, PID control system and the fuzzy neural network are also adopted respectively. The results of simulation prove the superiority of CFNN, which has the advantage of shortening training time and increasing control precision etc.