通过对电阻层析成像数据采集原理和深度学习网络的研究，提出了一种基于阵列电阻值和多层感知器深度学习网络相结合的流型识别方法。利用电阻层析成像系统中的16个电极传感器来获取流型样本数据，并构建出流型识别数据库，然后对多层感知器深度学习网络进行训练，获得可以识别不同流型的网络。实验结果表明，采用阵列电阻值结合多层感知器网络对流型进行学习和识别的方法，流型识别准确率可以达到95%，解决了流型图像生成过程与数据特征预选过程中流型特征损失的问题，流型识别性能得到了提高。

After studying the principle of resistance tomography data acquisition and deep learning network,a flow pattern recognition method based on array resistance value and multilayer perceptron depth learning network is proposed.The 16 electrode sensors in the electrical resistance tomography system are used to obtain the flow sample data,then the flow identification database is constructed,and the multilayer perceptron deep learning network is trained to obtain the network identifying different flow pattern.The experimental results show that the method of learning and recognizing the flow pattern by using the array resistance value and combining with the multilayer perceptron network can achieve 95% accuracy,which solves the problem of flow pattern loss during the process of flow pattern generation and data feature preselection,and the recognition performance has been improved.

TP183

四川省科技计划项目(2017JY0047)