基于数据驱动的卷积神经网络电容层析成像图像重建

Data-driven image reconstruction of electrical capacitance tomography based on convolutional neural network

针对电容层析成像技术的图像重建问题,提出了基于数据驱动的卷积神经网络图像重建方法。根据气固两相流的流型特点,通过数值模拟的方法随机生成了60000组介质分布图像,并利用有限元法计算了与之对应的电容向量,从而建立了一个"电容向量-介质分布"数据集;然后根据电容层析成像图像重建特点建立了卷积神经网络模型,对数据集中的训练集进行学习和训练,并利用测试集对训练结果进行了验证与评价。在此基础上,对获得的ECT图像重建卷积神经网络模型进行了静态实验和流化床测试实验研究。模拟和实验结果表明:所建立的卷积神经网络能较好地实现ECT图像重建,可直接用于流化床内的颗粒浓度分布测量。

A data-driven image reconstruction method based on convolutional neural networks is proposed for electrical capacitance tomography(ECT). According to the characteristics of the flow patterns of gas-solid twophase flow, 60000 sets of particle distribution images are randomly generated by numerical simulation and the corresponding capacitance vectors are calculated by the finite element method, thereby creating a "capacitance vector-particle distribution" dataset. Then a convolutional neural network model is developed to learn and train the training dataset. The training result is verified and evaluated with the testing dataset. Further, static experiments and fluidized bed measurement experiments are performed on the ECT image reconstruction with the obtained convolutional neural network model. Simulation and experimental results show that the established convolutional neural network can well reconstruct ECT images and can be directly used for particle concentration distribution measurement in a fluidized bed.