基于感知器神经网络的金属磁记忆检测管道缺陷分析

Analysis of Pipeline Defects by Metal Magnetic Memory Detection Based on Perceptron Neural Network

感知器神经网络可以在采用金属磁记忆技术查找管道隐性损伤的基础上,有效识别应力集中和宏观裂纹。对4项线性指标的感知器神经网络的计算机仿真分析,100次模拟的平均诊断正确率为71.2%。增加切向梯度和法向梯度乘积项的感知器神经网络识别效果最好,其100次模拟的平均诊断正确率达到了90.7%,显著高于线性模型的识别效果,可有效应用于金属磁记忆的管道缺陷监测。

The stress concentration and macroscopic crack between could be effectively distinguished by the perceptron neural network, on basis of hidden pipeline damages found by using the technology of metal magnetic memory. The average diagnostic accu- racy rate of 100 times of computer simulation analysis was reached 71.2% via perceptron neural network by 4 linear indexes. When adding the product of tangential gradient and normal gradient of perceptron neural network, the distinguish effect was at optimal, and the average diagnostic accuracy rate of 100 simulations was reached 90. 7%, which is significantly higher than that of the linear model, so it can be used effectively to detect the pipeline defects of metal magnetic memory.