磁信号检测高温合金内外壁缺陷分类研究
Research on the classification of inner and outer wall defects of superalloy based on magnetic signal detection
罗炜韬 1胡博 1石文泽 1王少飞 1陈宇 1樊梦 1程虹之1
作者信息
- 1. 南昌航空大学无损检测技术教育部重点实验室,江西南昌 330063
- 折叠
摘要
为区分高温合金内外壁缺陷,采用高精度弱磁传感器对高温合金人工槽型缺陷进行检测.从有缺陷的一面检测为正面(即外壁缺陷),从另一面检测为反面(相当于内壁缺陷),发现单侧表面缺陷检测信号出现极性相反的情况,无法通过极性相反这一特征区分内外壁缺陷.针对这一现象,基于支持向量机中的分类功能,对采集到的弱磁信号进行特征提取获得面积、幅值和占宽 3种特征量建立分类数据库,进行内外壁缺陷分类识别,并采用不同的核函数建立分类模型进行比较,最终发现径向基核函数分类模型正确率最高,为 84.37%,且经过样本库外的试件验证后正确率仍有 81.25%.结果表明该算法能够有效地对高温合金缺陷信号进行分析和辨识,具有一定的实用价值.
Abstract
In order to distinguish the internal defects and external defects of superalloys,high-precision weak magnetic probes are used to detect artificial groove defects of superalloys,detected as positive from the defective side(external defect)and negative from the other side(internal defect),it is found that the polarity of the detection signal of one side surface defect is opposite,and the inner and outer wall defects cannot be distinguished by the feature of opposite polarity.In response to this phenomenon,based on the classification function in the support vector machine,the weak magnetic signals collected are extracted by feature extraction to obtain three kinds of characteristic quantities of area,amplitude and width to establish a classification database,and then classify and identify the inner and outer wall defects.For comparison,the classification model established by the radial basis kernel function is finally found to have the highest accuracy rate,which is 84.37%,and the accuracy rate is still 81.25%after verification by the samples outside the sample library.The results show that the algorithm can effectively analyze and identify the defect signal of superalloy,and has certain practical value.
关键词
弱磁检测/高温合金/支持向量机/内外壁缺陷识别Key words
weak magnetic detection/superalloy/support vector machine/inner and outer wall defect identification引用本文复制引用
基金项目
国家自然科学基金(51967014)
江西省主要学科学术和技术带头人培养计划--青年人才项目(20204BCJ23001)
江西省研究生教育创新计划项目(YC2021-067)
出版年
2024
NETL
NSTL
万方数据