回归分析在漏磁定量检测中的应用

Application of Regression Analysis in Quantitative MFL Detecting

通过分析漏磁场，选取法向分量的峰峰值和峰峰值间距作为漏磁通信号的特征量。重点介绍了多元回归分析模型理论、参数估计方法和回归模型的假设检验，得到缺陷宽度、深度与峰峰值和峰峰值间距之间的回归模型，最后利用漏磁数据对回归模型进行了检验，结果表明预测缺陷宽度和深度的误差不超过10％，利用回归分析能够实现漏磁检测的定量化。

Through analysis of magnetic leakage field, the peak - peak values of the magnetic leakage signal normal component and the distance between the two peaks are selected as its feature index. This paper mainly presentes the theory of multiple regression analytical model and parameter estimation method and hypothesis test of regression model. The regression model is obtained which involves the relations of width, depth, peak -peak values and the distance of peak -peak values. The regression model is tested by the data of magnetic leakage, and the result shows that the errors of width and depth of the defect predicted are not more than 10%. This indicates that the regression analysis is a good method to realize the quantitative of magnetic flux leakage detecting.