The Research On The Intelligent Control Strategy Of Fuzzy-Based Neural Network

Today, intelligent control research areas, one of the major hot spots for the fuzzy neural network. The fuzzy neural network fuzzy neural network technology and the advantages of rolled into one, including the self-learning, adaptive, Lenovo and fuzzy information processing. It has a neural network learning ability and adaptive capacity of the network increased, while taking advantage of existing expertise, it has a strong reasoning ability. By combining their strengths, significantly enhanced the system's learning and expression.Adaptive is a fuzzy neural network has the advantage, however, the characteristics of the input data to a certain extent, constrained neural network control system, fuzzy neural structure and performance. For the simple structure of the control system, experienced a great amount of input data, a higher dimension, you tend to result in the training time is very long, and can not get a good convergence, or even convergence. Enter the amount of data in the small dimension is not high to adopt the structure of complex control systems, will result in computational speed is relatively long, and slow convergence.To address the above problems, this in-depth study of the fuzzy neural network structure and parameters learning algorithm based on the amount of input data characteristics and system performance requirements, using the appropriate strategies. In the small amount of input data and system requirements on the convergence and accuracy is not very high case, the use of the advantages of a simple neural network structure presents a fuzzy BP neural network model, BP algorithm is fully considered the initial over-reliance on the network start value of the problem, the input data to quantify the fuzzy processing, and then used in BP neural network model. At the same time the input dimension greater.In response to these established structural model, choose a different sample data control object to its simulation experiments carried out by simulation results show the feasibility and effectiveness of the method.