| This paper presents several control methods for the MOCVD device control system set to handle the non-linearity, time variability and large delay of plant controls, which include fuzzy control, neural network control, immune control and other compound control ones.Firstly, A kind of Fuzzy temperature controller is proposed in this paper. Fuzzy controller can realize the PID parameters self-regulation on-line. Secondly, as a common method to train multilayer perception networks, the standard BP algorithm, is lack of efficiency and adaptive capacity. Then a fast varied step and conjugate based on BP algorithm is presented. Thirdly, based on the feedback mechanism of the biologic system a compound control method that the conventional controller is combined with an immune controller is presented. In the concrete the immune controller is in series with the controller whose parameters have been optimized.In addition, based on the immune feedback regulating law and the approaching ability for nonlinear function of fuzzy rational logic, a fuzzy immune control is presented. Experiments show architectures designed in this dissertation are very valuable.Finally, with the development and wide spread of ANN's applying, there continually appears some problems, there exists a conflict between the network complexity and its generalization and so on. So based on introducing the performance of the RBF network on many aspects, the possibility and the validity of using an RBF network with the immune strategy as its learning algorithm to temperature control system is studied.In summary, there have been some research results in this paper about algorithm models based on intelligent control, which are innovative. On the other hand, there are also some explorations about applying these models in practice, and some good results or experience are received. Consummating the above models and exploring the applying area is the direction of our further research. |
PDF Full Text Request