A Fuzzy Neural Network And Its Application Of Rotary Kiln's Control

Fuzzy control and neural network have their own advantages and disadvantages, Fuzzy neural network combine the two technology together, using their advantages and avoiding their shortcomings. In other words, to neural networks, its nodes and the corresponding weights will be of a certain physical meaning by introducing the image thinking and reasoning of fuzzy systems. And to fuzzy system, its self-learning and self-adaptive capacity will be effectively enhanced by introducing the learning and computing capacity of neural network.We studied the typical fuzzy neural network, Mamdani model and T-S model firstly. Then we designed a new type of fuzzy neural network in this article, which was based on Average model. In the process of training the fuzzy neural network, we used the traditional gradient learning. However, the system showed local minimum problem, the performance was instable. Then we used the combination of gradient descent and simulated annealing to train the network, as the simulated annealing can flee the local minimum, so the system can achieve global minimum, the performance showed that this method was very useful. Finally, we used this new fuzzy neural network to control the complex rotary kiln system, and achieved good results.In this paper, the designed fuzzy neural network not only had the adaptive, self-learning ability of neural network, but also has the physical meaning of fuzzy systems. At the same time, the combination of gradient descent and simulated annealing algorithm solved the local minimum to some extent, which was also of certain practical significance.