基于UMAC的RBF神经网络PID控制

针对通用电机运动控制器（UMAC）下的传统PID控制和现有的模糊PID控制自适应性和鲁棒性较差，伺服系统的动静态性能不理想的问题，将RBF神经网络引入到UMAC的PID参数调节中，增强伺服系统的自适应性和鲁棒性，并提高系统动静态特性。通过UMAC的嵌入式PLC程序对算法进行了实现，位置阶跃响应实验和正弦跟踪实验表明，RBF神经网络PID控制下的伺服电机位置阶跃响应上升时间由传统PID控制下的0.164 s和模糊PID控制下的0.118 s减小到了0.017 s，峰值时间由传统PID控制下的0.196 s和模糊PID控制下的0.131 s减小到了0.023 s，调节时间由传统PID控制下的0.216 s和模糊PID控制下的0.142 s减小到了0.025 s，电机响应速度变快；RBF神经网络PID控制下的伺服电机位置正弦响应动态跟随最大误差由传统PID控制下的188 counts和模糊PID控制下的120 counts减小到了39 counts，且误差波动较小、平稳，伺服电机动态跟随性能显著提高。

RBF neural network tuning PID control based on UMAC

The self-adaptability and robustness of traditional PID control and current fuzzy-PID control adopted by universal motion and automation controller (UMAC) were not strong, and the static-dynamic performance of servosystem controlled by them was not ideal. In this paper, RBF neural network was adopted to automatically adjust PID control parameters, which could strengthen the self-adaptability and robustness of servosystem and improve the controlling characteristics of servo system. This control algorithm was implemented by embedded PLC program of UMAC. The experimental results of step response and sinusoidal tracking response show that the rise time of servo motorposition step response by RBF neural network tuning PID control decreases from 0.164 s by traditional PID control and 0.118 s by fuzzy-PID control to 0.017 s, the peak time decreases from 0.196 s by traditional PID control and 0.131 s by fuzzy-PID control to 0.023 s, and the setting time decreases from 0.216 s by traditional PID control and 0.142 s by fuzzy-PID control to 0.025 s, which mean that the motor responds faster. Meantime, the dynamic following error peak value of motor position sinusoidal response by RBF neural network tuning PID control decreases from 188 counts by traditional PID control and 120 counts by fuzzy-PID control to 39 counts, and the error fluctuation issmall and steady, which mean that the dynamic tracking performance of the motor is significantly improved.