厚管壁电容层析成像图像重建算法

Image reconstruction algorithms for electrical capacitance tomography with thick-wall pipeline

探讨了适合厚管壁条件下的电容层析成像图像重建算法.针对厚管壁管道内几种不同流型,分别采用LBP算法、Landweber迭代算法和BP神经网络对8电极电容传感器进行成像重建计算.结果表明:在厚管壁情况下,LBP算法重建的图像质量很差;Landweber迭代算法在层流下的重建效果好于核心流和环状流;而BP神经网络算法可以有效重建管道内的介质分布,但对于没有训练到的任意流型,其重建效果不够理想.

An ECT （electrical capacitance tomography ） sensor with thick-wall pipeline is researched, including sensitivity map characteristic and suitable image reconstruction algorithms. For different flow distribution patterns in thick-wall pipeline, the algorithms such as linear back-projection （LBP）, Landweber iteration and BP neural networks are adopted to reconstruct the images in 8- electrode ECT sensor. The reconstruction results demonstrate that the reconstructed image quality with LBP is poor. Meanwhile, the reconstructed image in stratified flow is better than the ones in core flow and annular flow with Landweber algorithm. More over, BP neural networks algorithm is the only effective way to reconstruct the inner-pipe permittivity distribution, but the result is not satisfactory when the flow distribution patterns are not trained.