Gas-Liquid Two-Phase Flow Parameter Measurement Method Based On Deep Learning

Two-phase flow measurement is widely used in industrial production processes,and its measurement accuracy plays a vital role in the economic stable operation of these industries.Due to the diversity and complexity,the measurement of two-phase flow parameters has always been a difficult problem in this field.Based on the resistance tomography technology and deep learning network,a two-phase flow pattern recognition and flow measurement method is proposed,which overcomes the problem of feature loss in traditional tomography,improves the flow pattern recognition rate and flow accuracy.In this paper,the effectiveness of the original unprocessed and individually unexplainable feature extraction is improved by increasing the number of hidden layers of the multilayer perceptron,and a gas-liquid two-phase flow pattern recognition system is established based on the multilayer perceptron.The 120 groups of raw resistance data collected by the 16-electrode sensor in the tomography system were used as the system's input signals to train a multi-layer perceptron network using30,000 samples.The successfully trained a flow pattern recognition system,which overcomes the problem of flow pattern feature loss during traditional tomographic flow pattern image reconstruction and data feature extraction;Then,based on the results of flow pattern recognition,120 groups of real-time collected data from resistance tomography are used as the input of the long short-term memory network,which can well deal with the long-distance dependencies.Thus,a flow measurement system based on a long short-term memory network is established,and the measurement of the bullet shape flow and bubble flow is realized.Experimental results show that the flow pattern recognition system based on multi-layer perceptron can make the accuracy of flow pattern recognition reach 95%;The liquid flow measurement method based on long short-term memory networks can make the measurement accuracy of liquid flow within 3%,which basically meets the requirements of industrial measurement.