基于神经网络PID的丝杠运动误差控制

Control of screw motion error based on neural network PID

针对丝杠传动系统具有非线性及缓时变的特点,使用了BP神经网络PID控制器,该控制策略能在线调整PID控制参数.为了改善常规BP神经网络收敛性慢和系统不稳定的现象,提出了误差函数在线调整学习率的方法,建立了丝杠传动系统被控对象的数学模型,并给出了具体的控制算法.在先进制造过程对象试验平台(AMPOT)上进行了丝杠运动误差补偿试验,试验结果表明,该控制方法可将丝杠的运动误差减小50%～60%.

Aiming at the nonlinearity and slow time-variant in screw drive system, a control- ler of Neural Network PID that can adjust parameters of PID on-line is used. In order to im- prove the convergence slow and system unstable phenomena of conventional BP Neural net- work, the method using the error function to adjust the learning rate online is adopted. The controlled object mathematical model of screw drive system is established and the detailed control algorithm was put forward. The experiment of error compensation on the advanced manufacturing process object platform （AMPOT） is carried out, the result of experiment show that the control method can reduce the screw error by 50 %- 60%.