Application of Multiple Models BP NN Weighting Optimal Controller in medium plate cooling process

In the controlled cooling process, the medium plate controlled usually contains some large scale of uncertainty on thickness. So, if only one model is utilized constantly, it can not get the satisfied precision when the thickness of the plate differs largely from the real one. In this paper, a novel Weighting Multiple Models Controller (WMMC) using multiple BP Neural Networks (BP NN) is proposed and applied. Firstly, multiple models are established according to the traits of the thickness of the plates under certain circumstance. Then the weighting parameter can be forecasted by one BP NN according to thickness zone, which is trained online, to establish the process model. Lastly, the optimal controller is presented according to the above model. Utilizing the choice of the polynomials, it not only eliminates the steady state error, but also places the poles of the closed loop system arbitrarily. The result shows that the proposed WMMC is superior to the conventional controller in plate cooling control and has wide application prospect.