管道金属磁记忆检测缺陷的判别分析模型

Discriminant Analysis Model for Defect Recognition of Metal Pipeline Magnetic Memory Detection

线性判别分析模型能较好地识别应力集中和宏观裂纹这两种缺陷,通过计算机仿真分析,100次模拟的平均诊断正确率达到了85.5%。经过改进的带切向梯度和法向梯度乘积的非线性判别模型的识别效果最好,其100次模拟的平均诊断正确率达到了91.4%,显著高于线性模型的识别效果。

Linear diseriminant analysis can easily distinguish two-type defects, and the average accuracy rate of 100 computer simulations reached 85.5%. The simulation result shows that the improved non-linear diseriminant analysis which boasting of tangential gradient and normal gradient outperforms the linear model in defeet recognition because of a 91.4% accuracy rate.