导电结构亚表面缺陷电涡流/超声复合检测方法

摘要

An integrative nondestructive testing (NDT) method based on the eddy current testing (ECT) and ultrasound testing (UT) was analyzed for the problem of limited detection accuracy and range of single detection method for conductive structure sub-surface defects inspection.The techniques for calculating the ECT/UT basic probability distribution, the distribution of weights of evidence and the evidence fusion to obtain accurate identification results were investigated in order to solve the problem of heterogeneous evidence confliction and to fully exploit the advantages of individual modality detection method.A data fusion method based on fitting variance optimized D-S evidence theory was proposed.A normal fitting method based on the Akaike information criterion (AIC) was conducted to gain the basic probability distribution curve of ECT/UT data in order to reduce fitting errors and prevent over-fitting problems.The Bayesian criteria were used to calculate the posterior probability according to the applicability of ECT/UT in the detection.The fitting variance was used for evidence weight optimization.The joint probability density distribution was obtained by D-S evidence theory.The ECT/UT integrative test result was calculated.%针对导电结构体亚表面缺陷检测中单一检测方法准确率和检测范围有限等问题,研究电涡流与超声复合无损检测(NDT)方法.为了解决异源传感器复合过程中的证据冲突问题,充分发挥复合检测方法的优势,研究获取电涡流/超声检测基本概率分布、证据权重分配、证据融合以获得准确识别结果的技术,提出基于拟合方差的D-S证据理论的复合检测方法.为了降低拟合误差,防止过拟合,利用基于赤池信息量准则的正态拟合方法获得电涡流检测和超声检测的概率分布,用贝叶斯准则求取后验概率.根据电涡流和超声在检测中的适用性,以拟合方差对证据权重进行优化,利用D-S证据理论对后验概率密度进行融合,获得联合概率密度分布,求取电涡流/超声复合检测结果.