孔洞缺陷的红外无损检测和PNN识别与定量评估

发布时间：2018-11-05 07:20

【摘要】：针对红外无损检测中因特征信息缺失,致使识别与评估效果不佳这一问题,研究以铝板为对象,基于红外无损检测技术,结合主成分分析和概率神经网络对铝板正常区及三类孔洞缺陷区进行了识别与面积定量评估。研究首先采集铝板降温过程的红外时序热图,提取了正常区和各类孔洞缺陷区的时序灰度值作为初始特征。其次,采用主成分分析对初始特征进行提取,并结合概率神经网络,以像素点为单位实现孔洞缺陷的识别及面积定量评估,并采用了支持向量机进行了对比研究。实验结果表明,对于正常区和三类孔洞缺陷区测试样本的面积评估正确率分别为99.6%、97.0%、94.7%和93.0%,相比支持向量机的评估结果,所提出的研究方法具有更高的正确率。研究论证了采用主成分分析和概率神经网络,基于时序特征,以像素点为单位,实现孔洞缺陷识别和面积定量分析的有效性和准确性。
[Abstract]:In order to solve the problem that the recognition and evaluation effect is not good due to the lack of feature information in infrared nondestructive testing (NDT), this paper studies the technology of infrared nondestructive testing (NDT) based on the aluminum plate as the object. Combined with principal component analysis (PCA) and probabilistic neural network (PNN), the normal and three kinds of hole defect areas of aluminum plates were identified and quantitatively evaluated. In this paper, the infrared time series thermal diagram of aluminum plate cooling process is first collected, and the time series gray values of the normal region and various hole defect regions are extracted as the initial characteristics. Secondly, principal component analysis (PCA) is used to extract the initial feature, combining with probabilistic neural network, the hole defect recognition and area quantitative evaluation are realized in pixel units, and the support vector machine (SVM) is used to carry out a comparative study. The experimental results show that the correct rate of area assessment for normal and three kinds of hole defect areas is 99.6% and 93.0% respectively. Compared with the results of support vector machine, the accuracy of the test samples is 99.6% and 93.0% respectively. The proposed method has higher accuracy. The effectiveness and accuracy of using principal component analysis and probabilistic neural network to realize hole defect identification and area quantitative analysis are demonstrated based on temporal features and pixel points.
【作者单位】：华东交通大学机电工程学院;
【基金】：国家自然科学基金项目(51175175) 江西省教育厅科技项目(GJJ13342)
【分类号】：TN219;TG115.28

【参考文献】