关键词: 参数优化, 反向传播算法, 机械加工, 模糊逻辑, 误差复映, 自适应模糊神经推理系统

Abstract: Training arithmetic of adaptive network-based fuzzy Inference system (ANFIS) is improved with conjugate gradient algorithm on the basis of analyses of common improving methods of back propagation algorithm. During training Fletcher-Reeves method is used to compute influence factor of last search direction to new search direction. It’s proved that it cost fewer iterations and time to converge with improved arithmetic than standard ANFIS arithmetic by applications in chaotic time-series prediction and approaching complex non-linear functions. How to improve arithmetic on the basis of standard arithmetic with fuzzy toolbox is enlarged on. At present development of computer aided process plan is slowed by many complex non-linear problems. In order to utilize the learning, adaptive and logic inference abilities of ANFIS to solve them, improved arithmetic is used to optimize machining parameters by approaching the non-linear relationship among error reflection coefficient and rigidity of machining system, feeding speed etc. In this way work efficiency and adaptability of machining system are improved. Feasibility of this method is validated by experiments.

Key words: Adaptive network-based fuzzy inference system(ANFIS), Back propagation, Error reflection, Fuzzy logic, Machining, Parameters optimization