大数据环境下超声波焊缝缺陷识别方法的研究

发布时间：2018-01-14 06:01

本文关键词：大数据环境下超声波焊缝缺陷识别方法的研究　出处：《计算机工程与应用》2016年23期 　论文类型：期刊论文

【摘要】：为了解决常规超声波焊缝缺陷识别方法分类模型固定和训练集规模有限而难以体现不同缺陷的差异性和同类缺陷的多态性的问题,结合当今大数据环境下的数据分析策略和基因缺陷识别中匹配的思想,通过主成分分析和CURE聚类算法将缺陷回波信号编码转换成可进行匹配的对象,进而将当前检测缺陷特征与历史检测数据进行匹配,并利用最近邻方法实现了对缺陷历史检测数据集的扩充。通过在R上应用基于基本空位罚分的Smith-Waterman比对算法进行仿真实验验证了该缺陷识别方法是可行的,有效地识别了气孔、夹渣、裂纹、未焊透和未熔合五类常见缺陷,具有较好的识别准确率。
[Abstract]:In order to solve the problem that the classification model and training set size of conventional ultrasonic seam defect identification method are fixed and limited, it is difficult to reflect the difference of different defects and the polymorphism of the same kind of defects. Combined with the strategy of data analysis in large data environment and the idea of matching in gene defect recognition, the defect echo signal coding is transformed into a matching object by principal component analysis and CURE clustering algorithm. Furthermore, the current defect detection feature is matched with the historical detection data. The extension of defect history detection data set is realized by using nearest neighbor method, and the Smith-Waterman matching algorithm based on basic vacancy penalty points is applied to R to verify the deficiency. The method of trap recognition is feasible. Five kinds of common defects, such as porosity, slag inclusion, crack, incomplete penetration and incomplete fusion, are effectively identified, and the recognition accuracy is good.
【作者单位】：东北林业大学信息与计算机工程学院;
【基金】：国家自然科学基金(No.31370565,No.61300098) 哈尔滨市人才项目专项(No.2015RAYXJ005)
【分类号】：TG441.7;TP391.41
【正文快照】： 1引言目前,常规的焊缝缺陷检测主要利用小波包分解或分析超声回波信号的能量变化来提取缺陷特征,用支持向量机或图像分割方法来构建缺陷识别分类器,在理论和应用中都取得了一定的成果[1-5]。这些方法一般是构建固定的缺陷分类模型并且基于数据量有限的训练集,但超声波焊缝缺陷

【相似文献】