基于RBF辨识的模糊神经网络控制器的设计与实现

Design and Realization of Fuzzy Neural Network Controller Based on RBF Identification

随着众多新型模糊神经网络被提出,针对模糊神经网络具有的典型特点,即需要对输入输出数据范围进行转化和处理,所涉及到的对量化因子和比例因子的实时调节问题,该文提出一种优化方案。其依据神经网络具有的自学习能力,通过增加模糊神经网络的层数,提出一种包含对量化因子和比例因子调节的改进型模糊神经网络,以减少系统的辅助优化环节。同时,引入辨识性能较好的径向基函数神经网络(RBF)为系统提供精确的Jacob ian信息,取代常规的近似做法。最后结合实例仿真证明了该优化方案的合理性。

With many new fuzzy neural networks proposed, this paper gives an optimized method aiming at the typical trait of fuzzy neural network which needs dealing with the range of the input and output. It refers to adjusting the scaling factor for fuzzification and the scaling gain in real time. Based on the neural network＇s capabilities in self - learning, the improved fuzzy neural network can optimize the scaling factor for fuzzification and the scaling gain and remove the assistant optimization part by increasing the layers. Besides, the radial basis function （RBF） neural network that has good capabilities in identification is introduced to offer the precise information of Jacobian for the system. Finally, the feasibility of this new method has been proven by the simulation.