Internal model control based on self-constructing wavelet neural network

In this paper, a novel internal model control (IMC) algorithm based on self-constructing wavelet neural network (WNN) is proposed for the nonlinear process. The self-constructing learning algorithm is composed of the structure learning and the parameters learning. In the structure learning phase, the similarity measurement method is adopted to determine whether or not to add a new wavelet base to satisfy the identification requirement. In addition, the impact of wavelet base on the output of the network is used as the basis for deciding whether to cut the wavelet base. The gradient descent method which can adjust the learning rate automatically is applied in the parameters learning. Combining the self-constructing WNN with IMC, the networks used to identify the process model and the controller model can dynamically determine the number of nodes. So, the networks convergence speed could be increased, and the dynamic performance and robustness of the system could be improved.