新型ANDOR模糊神经网络的研讨.pdf

th Proceedings of the 25 Chinese Control Conference 7-11 August, 2006, Harbin, Heilongjiang 新型AND-OR模糊神经网络的研究 隋江华，任光 大连海事大学轮机工程学院, 大连 116026 E-mail: suijh@, reng@ 摘 要:本文介绍了一种新型的AND-OR模糊神经网络，推导出每一层的输入输出映射关系式，并证明AND-OR 模糊神经网络在一定条件下其功能等价于模糊推理。利用神经网络结构的遗传优化自动提取模糊规则，并设计 船舶运动AND-OR模糊自适应控制器，与传统模糊控制系统相比不仅具有缩小输入空间，降低模糊规则数的优 点，而且仿真效果也令人满意。 关键词:模糊AND-OR神经网络，入度，相关度，连通度 The Novel AND-OR Fuzzy Neural Network Sui Jianghua, Ren Guang Marine Engineering College, Dalian Maritime University, Dalian, 116026 E-mail: suijh@, reng@ Abstract: A novel AND-OR Fuzzy neural network is introduced and the symbol expression is educed in every layer. The equivalence between fuzzy inference and AND-OR FNN is proved. An approach to auto-extracting fuzzy rules is obtained by training the structure of the AND-OR FNN. The AND-OR FNN adaptive controller for ship control is designed, whose superiority is shown by comparing to the conventional fuzzy control system, Though the input space is reduced and the number of rules is decreased, the simulation results illustrate the approach is practicable, simple and effective and the performance index is much better. Key Words: AND-OR FNN,In-degree, Relevant degree, Connectivity 糊神经元（OWA ）[5]。J.Kim等人用遗传算法来优 1. 引 言(Introduction) 化设计模糊前馈神经网络[6]。J.J.Buckley研究神经 网络、连续函数和模糊系统之间的数值关系，证明 模糊控制理论已经成功地应用到很多领域，然而 了混合神经网络能实现连续模糊函数。 知识获取的“瓶颈”问题和输入维数的“爆炸”问 W.Pedrycz 是