Recognizing Flow Pattern of Gas/Liquid Two-component Flow Using Fuzzy Logical Neural Network

摘要

This paper describes a new method based on fuzzy neural network which is used to recognize the two-component flow pattern. The paper discusses the structure of the fuzzy neural network, including the selection of the fuzzy logical rule and the training sets. An accelerated learning algorithm (adaptive backward propagation algorithm) is used to train the neural network to shorten its learning time. After computer simulation has been done, it is found that this new method can recognize four typical flow patterns existing in the gas/liquid two-component flow, which are stratified flow, annular flow, slug flow and buble flow. Finally, Some results useful for the future work are also presented.