Abstract: Aiming at improvement of image reconstruction algorithm in 12-Electrical Capacitance Tomography system，this paper conducts an experimental study on closed containers gas/solid two phase flow，which mainly depend on the stability and the speed of image reconstruction algorithm.To keep the stability of solving process and the computational performance，the image reconstruction algorithm based on a new set of neural network types supported algorithm（NSSN） is first applied to the ECT image reconstruction system.Large-scale training of neural network algorithm puts forward a slow sub-division of the network enhancement.The system uses 12-electrode capacitance tomography system for gas-solid flow tube closure data detection，the use of the improved neural network algorithm for image reconstruction.The obtained experimental results fully show that the image reconstruction method has high precision imaging，computing speed and so on.This method can simplify the use of neural network structure，reducing the size of neurons for image reconstruction of electrical capacitance tomography system and providing a new way of thinking.

Key words: Electrical Capacitance Tomography（ECT）, New Supported Set Network（NSSN）, dividing subnetwork, image reconstruction

摘要： 以12电极电容阵列传感器ECT系统为背景，从图像重建的稳定性和速度两方面对密闭容器中气-固两相流场的图像重建算法优化进行实验室研究。将基于新型类支集函数的神经网络算法（NSSN），应用于ECT系统图像重建算法中，使得图像重建算法的求解过程稳定并具有良好的计算性能。针对大规模神经网络算法训练速度较慢的问题提出了划分子网络的改进方法。通过对封闭管道的气固两相流进行数据检测，并采用改进后的神经网络算法进行图像重建，实验结果验证了改进后的方法弥补了大规模神经网络运算速度慢的不足，可以简化神经网络的结构，减少神经元的规模，为电容层析成像系统图像重建提供了新的思路。

关键词: 电容层析成像, 新型类支集神经网络, 划分子网络, 图像重建