论文部分内容阅读
基于主成分分析-支持向量机(PCA-SVM)模型,提出一种利用近红外辐射信号预测接头形貌的方法,研究了信号的变化规律与焊缝形貌之间的相关性,实现了工艺参数的优化。提取信号的6种时域特征参数并进行主成分分析,获得了接头形貌综合评定指标。根据信号的输入特征,利用支持向量机进行了分类预测。结果表明,近红外辐射信号能够反映焊接过程中焊缝状态的变化,不同缺陷的特征变化具有较大差异,且存在清晰的识别度。该预测模型能够准确识别焊缝成形形貌,准确率高达96.6%。
Based on principal component analysis and support vector machine (PCA-SVM) model, a method of predicting the shape of the joint by using near-infrared radiation signals was proposed. The correlation between the signal variation and the morphology of the weld was studied, Parameter optimization. Six kinds of time-domain characteristic parameters of the signal were extracted and analyzed by principal component analysis. The comprehensive evaluation index of the joint morphology was obtained. According to the input characteristics of the signal, using SVM classifier prediction. The results show that the near-infrared radiation signal can reflect the change of the weld state during the welding process, the characteristics of different defects have great differences, and there is a clear recognition degree. The prediction model can accurately identify the weld profile, the accuracy rate as high as 96.6%.