Research and Application on Two-stage Fuzzy Neural Network Temperature Control System for Industrial Heating Furnace

Abstract

Industrial heating furnace has a great deal of special characteristics such as big capacity, long lag and non-linear trait, etc. In order to control it better, we present a sort of fuzzy neural network temperature control model. It can transform the rulers of fuzzy logic control to a pair of input-output samples of multilayer forward neural network.The knowledge is not expressed by a serial of rules but distributed into the whole network. Based on this model, we have designed a two-stage fuzzy neural network temperature control system for industrial heating furnace. The first-stage controller is responsible for determining the control variable according to deviation information of controlled variable. The second-stage controller takes charges of adjusting the control variable coming from the first-stage controller through other process parameters. The system takes full account of the impact of many process parameters on controlled variable. It uses two-stage fuzzy neural network controller to decentralize process of control parameters, which makes it easy to extract fuzzy rules, greatly reduces the number of fuzzy rules and produces reasonable control outputs. Engineering applications show that the system has a lot of advantages such as high accuracy, strong robustness, etc. Its quality is superior to conventional control and it is suitable for long lag, non-linear system in particular.