Grey prediction based RBF neural network self-tuning PID control for turning process

Turning process is a system with the property of nonlinear, time varying, incalculability and uncertainty. Conventional PID controller tuned at typical operating point can hardly work well at different operating condition. A novel self-tuning PID control strategy based on a two-level radial basis function (RBF) neural network (NN) is proposed, the two-level NN are called static NN (SNN) and dynamic NN (DNN) separately. SNN is used for controller PID arguments' primary tuning according to the system operating point such as turning depth of CNC, in order to follow the wide range turning depth changing; DNN is used for PID fine tuning according to the error and error rates, in order to overcome the small range turning depth changing, system parameters' slow time varying and some disturbance. For overcoming the inertia and time delay of the controlled plant, grey predictor is introduced to predict future output value. The predictive result is used as tuning information of DNN. Simulation results of a turning process show that good dynamic regulation performance can he obtained by using the presented new method, and stronger robustness is obtained.