作为一种新型的非接触式检测方法, 基于红外热成像技术的机载电路板故障模式诊断方法受到越来越多的关注。本文针对传统基于红外热图的电路板故障检测算法中存在的缺陷, 提出一种结合红外图像分割、热阻网络、支持向量机 SVM(Support Vector Machine)与 D-S证据理论的故障检测算法。首先, 通过红外图像分割完成目标芯片区域温度提取, 应用热阻网络模型对目标区域温度信息进行优化; 其次, 提取温度信息特征向量分别输入对应的初级 SVM诊断模块, 输出各故障模式的加权基本概率分配值 BPA(Basic Probability Assignment); 最后, 应用 D-S证据理论对各证据体加权 BPA进行数据融合, 输出融合后的故障诊断结果。实验结果表明, 本文算法加强了有效证据体对诊断结果的正面影响, 削弱了无效证据体的负面影响, 大幅度提高了机载电路板故障模式诊断准确度。

Infrared imaging technology is a new non-contact detection method, which has received attention in the field of on-board circuit board fault diagnosis. This paper combines image segmentation, thermal resistance network, SVM and D-S evidence theory to solve the problems in the traditional circuit board fault detection algorithms. Firstly, Image segmentation and thermal resistance networks are used to complete target region temperature extraction and optimization. Secondly, the temperature feature vector is input to the SVM to obtain a weighted BPA. Finally, D-S evidence theory is used to complete data fusion and obtain fault diagnosis results. The experimental results show that the proposed algorithm enhances the positive impact of effective evidence on the diagnosis results, and greatly improves the diagnostic accuracy of the circuit board failure mode.