基于改进BP算法神经网络的超声波电动机速度控制

针对超声波电动机没有精确数学模型,输出具有很强的时变性和非线性的特点,提出了一种改进BP神经网络控制器,神经网络由输入层、隶属函数层、规则层和输出层四层节点构成,在传统BP神经网络基础上,加入了模糊偏差单元和关联节点,使规则层不仅接收来自隶属函数层输出的信号,还接收自身的延时输出信号,能够存储过去的输入输出信息,提高控制系统学习记忆的稳定性。将改进BP神经网络控制器应用于行波超声波电动机速度控制,仿真实验验证了该方法的有效性,与传统BP神经网络相比较,控制精度、响应速度都有改善。

Neural Network Controller Based on Modified BP Algorithm for Ultrasonic Motors

Ultrasonic motor has no precise mathematical model presently,and its output has strong time-variation and nonlinearity.A modified BP neural network controller was presented in this paper which consisted of input layer,membership function layer,rule layer and output layer.Based on the traditional BP neural network theory,by adding fuzzy deviation neuron and relative neuron to the net,the rule layer received not only the signal from the membership function layer but also the delayed output signal of itself,so it can store the past input and output information to improve the stability of the study and memory.The traveling wave ultrasonic motor's revolving speed was controlled by using the modified BP neural network controller.The emulation results revealed the control accuracy and response speed improved compared with the traditional BP neural network.