用组合算法逐步提高电阻层析成像的分辨率  

Progressive enhancing the resolution of electrical resistance tomographic image using combinatorial inversion algorithm

作　　者：石天明[1] 张亮[1] 肖理庆[1] 袁少玲[1] 周德全[1] 

机构地区：[1]中国石油大学信息与控制工程学院,东营257061

出　　处：《仪器仪表学报》2008年第12期2587-2593,共7页Chinese Journal of Scientific Instrument

基　　金：中国石油大学211建设基金(211SZ-05)资助项目

摘　　要：电阻层析测量中的软场效应和反演算法中的不适定性,常使作为测量结果的两相流层析图在相界面处呈模糊状态,给分析带来困难。采用迭代的定量性优化算法(如改进牛顿-拉夫逊类MNR算法),其图像分辨率虽较非迭代简单反演算法(如线性反投影LBP算法、灵敏度算法)的结果有改进,但当流型稍微复杂一点尤其当界面纵横交错时,迭代误差可能在某一个局部最优的水平上停留。如直接采用能全局寻优的基于遗传原理的反演算法,因其实际收敛性受到个体解结构中基因数量限制,层析图像的分辨力也受到限制。根据这个问题,本文采用LBP-MNR-GA组合算法层析成像的方法,应用组合算法中各相关算法的各自优点,逐步提取来自测量数据中的信息量,在提高层析图像分辨率上取得了较好的结果。The existence of soft field effect and ill-poseness in electrical resistance tomographic measurement often causes a blurring tomogram, which makes the analysis of two-phase flow a difficult task. Using iteration based sophisticated quantitative algorithm （such as the modified Newton Raphson method） could yield a better image resolution compared with that obtained from a simpler qualitative based algorithm （ such as the linear back projection method, or the sensitivity method）. The iteration might still be trapped in a local optimum point when the flow regime goes complex due to random and criss-crossed interface. Direct employing so-called global optimized generic algorithm is practically restrained by the size of genes allowable in an individual GA solution, leading to a very limited image resolution. To address this problem, an LBP-MNR-GA based combinational inversion algorithm is adopted in this paper to extract the information from the measurement data for a two-phase flow in a separate and progressive manner, which yields a better image resolution.

关 键 词：电阻抗层析测量 组合反演算法 图像分辨率与评价 

分 类 号：TH701[机械工程—仪器科学与技术]