基于RBF神经网络和粒子群算法的ECT传感器结构优化

给出一种RBF神经网络与粒子群算法相结合的电容层析成像（ECT）传感器结构参数优化方法。该方法以敏感场整体灵敏度大小等系统性能为优化目标，基于管壁厚度、屏蔽层厚度、径向屏蔽插入管壁深度、径向电极宽度、电极宽度（中心夹角）、管壁材料的相对介电常数、屏蔽层填充物相对介电常数7种重要的结构参数进行试验。应用RBF神经网络对多组结构参数以及对应的系统性能指标进行学习，得到回归模型，并应用粒子群算法进行寻优。结果显示，该方法参数寻优范围大，局限性小，寻优过程收敛快。优化后的系统整体灵敏度增大，成像质量改进。

Sensors Structure Optimization of ECT Based on RBF Neural Network and PSO

A new method optimizing structure parameters of ECT sensors was brought up, which was the combination of RBF neural networks and PSO, and whose optimization objectives were the value of the overall sensitivity and other system performances. Performed tests based on seven kinds of important structural parameters which were wall thickness, shielding thickness, the radial wall shielding insertion depth, the width of the radial electrode, the electrode width （ central angle）, the relative dielectric constant of wall material and filler shield relative dielectric constant, and studied on structural parameters and the corresponding performances by RBF neural network, the optimization based on PSO was performed with the regression model. The results show that the method has wide range of the parameters to optimize, fewer limitations and fast convergence process. The system optimized increases the sensitivity value, and improves the quality of the image.