油气井套管形变缺陷的电涡流检测数据量化评估方法  

Quantitative Evaluation Method of Eddy Current Detection Data for Deformation Defects in Oil and Gas Well Casing

作　　者：徐菲 黄华[1] 罗庆 XU Fei;HUANG Hua;LUO Qing(Oil Engineering Technology Research Institute,Zhongyuan Oilfield Company,SINOPEC,Puyang,Henan 457001,China;College of Geopgysics and Petroleum Resources,Yangtze University,Wuhan,Hubei 430100,China)

机构地区：[1]中国石化中原油田分公司石油工程技术研究院,河南濮阳457000 [2]长江大学地球物理与石油资源学院,湖北武汉430100

出　　处：《测井技术》2021年第6期584-591,共8页Well Logging Technology

基　　金：中国石油化工集团有限公司“十条龙”科技攻关项目“高酸性气田长周期安全生产关键技术研究与应用”(320093-001)。

摘　　要：针对油气井出现套管单面挤压、弯曲变形等缺陷问题,仅对缺陷进行定性分析不能完全满足套管损伤检测需求。提出一种数据驱动检测油气井套管形变缺陷的电涡流检测数据量化评估方法。该方法根据电磁探伤仪测量得到的原始数据进行预处理并提取11个A探头电压作为特征,根据11个A探头信号与同深度最小内径的对应关系选择数据驱动的算法进行拟合建模,并对比多层全连接神经网络、最小二乘支持向量回归(LSSVR)算法的建模效果,验证了数据驱动方法的缺陷辨识能力。同时实验表明,LSSVR算法相较于多层全连接神经网络算法,具有更好的数据拟合准确度和泛化能力,能够解决油气井套管形变电涡流检测数据的建模分析问题,在该类工业数据量化评估问题中具有更高的应用价值。In view of defects such as single-side extrusion and bending deformation in oil and gas well casing,only qualitative analysis cannot fully meet the needs of casing damage detection.So a data-driven quantitative evaluation method of eddy current detection data for deformation defects in oil and gas well casing is proposed in this paper.The original data measured by the electromagnetic flaw detector is preprocessed and 11 A probe voltages are extracted as features characterizing defects.According to the relationship between 11 probe signals and the minimum inner diameter at the same depth,a data-driven algorithm is selected for modeling.In this paper,multi-layer fully connected neural network and least square support vector regression(LSSVR)algorithm are adopted relatively to model the data and performances of them are compared.According to the results of our experiments,the defect identification ability of the data-driven method is verified,and LSSVR algorithm is proved to have better data fitting accuracy and generalization ability than neural network,capable of dealing with the issue of modeling and analysis for eddy current detection data in oil and gas well casing deformation.It implies higher application value of LSSVR algorithm in the quantitative evaluation of such industrial data.

关 键 词：无损检测 油田套管缺陷 电涡流 多层全连接神经网络 最小二乘支持向量回归算法 

分 类 号：P631.84[天文地球—地质矿产勘探]