基于BP神经网络的N80钢CO_2腐蚀预测方法研究

Prediction of CO_2 corrosion in N80 steel based on BP artificial neural networks methods

采用灰度数据矩阵统计、小波变换和二值化等方法对N80钢CO2腐蚀图像进行特征提取。结合BP神经网络技术,以腐蚀图像的各向异性和小波变换后子图像的能量参数作为腐蚀类型判据,建立基于BP神经网络的孔蚀速率诊断模型,实现了CO2腐蚀类型和腐蚀程度的预测。诊断结果与实验结果较好吻合。

The carbon dioxide corrosion morphologies of NSO steel were extracted using grey level data matrix statistic, wavelet transform and image binarizing methods. In combination with the multiplayer feed forward neural networks, a pitting velocity diagnosis model was developed based on the anisotropic energy parameter of corrosion images and the image energy parameter after wavelet transform. The type and degree of carbon dioxide corrosion were forecasted based on this model. The diagnosis model agreed well with the testing results.