基于卷积神经网络与门控循环单元的气液两相流流型识别方法

提出了一种基于卷积神经网络(CNN)与门控循环单元(GRU)的垂直管道气液两相流流型识别方法。该方法基于电阻层析成像(ERT)系统的重建图像,对其填充处理后进行离散余弦变换(DCT),求取最大、最小 DCT 系数的差值,选取一定帧数长度数据作为网络输入,对流型进行识别。分析了输入序列长度对CNN-GRU、CNN 及 GRU 网络分类准确的影响,确定了最佳输入向量维度分别为 60、65 及 50,使用实验数据对3种网络进行训练、测试,结果表明,CNN-GRU网络分类准确率最高,平均流型识别准确率可达 99.40%。

Flow Pattern Recognition of Gas-liquid Two-phase Flow Based on Convolutional Neural Network and Gated Recurrent Unit

The method for identifying the flow pattern of gas-liquid two-phase flow in a vertical pipeline based on convolutional neural network (CNN) and gated recurrent unit (GRU) is presented in this paper. Based on the reconstructed image by the electrical resistance tomography (ERT) system, the discrete cosine transform (DCT) is performed after the filling processing. The difference between the maximum and minimum DCT coefficients is calculated, and a certain frame length data are selected as the network input to identify the flow pattern. The influence of the length of the input sequence on the accuracy of CNN-GRU, CNN and GRU network classification is analyzed, and the optimal input vector dimensions are determined to be 60, 65 and 50. Using experimental data to train and test the CNN-GRU network, and compare it with the GRU and CNN networks, the results show that the CNN-GRU network has the highest classification accuracy, and the average flow pattern recognition accuracy rate can reach 99.40%.