基于直接转矩控制的感应电动机模糊神经元网络实现

针对感应电动机直接转矩控制模糊实现中模糊规则的参数难以设计问题，文章提出了一种基直接转矩控制的模糊神经元网络实现方法，来自动调整模糊系统中设计参数。在直接转矩控制模糊实现中，由于模糊推理和化函数具有不可导性，文章提出了一种“软最大化”算子。并根据模糊实现及该算子，设计了一个6层模糊神经元网络，对该模糊神经元网络采用改进动态学习率的BP算法进行学习。仿真试验表明：利用学习过的模糊神经元网络可以改进系统的控制性能。

Fuzzy Neural Network Implementation of Direct Torque Control of Inductance Machine

Due to the parameters in fuzzy implementation rules based on direct torque control of inductance motor are difficult to determine,this paper proposes a fuzzy neural network implementation based on direct torque control. A method of 'soft maximum' is derived because first derivatives of fuzzy inference functions and defuzzification function don't exist. According to fuzzy control system and the 'soft maximum' method, a six layer fuzzy neural network is designed. Its weights are trained using the modified dynamic learning rate back propagation to accelerate the convergence.Simulation shows that it can get good performance.