神经网络在海水腐蚀预测中的应用

Application of Artificial Neural Network on the Forecast of Seawater Corrosion

根据我国材料自然环境腐蚀网站长期以来积累的海水腐蚀数据,采用BP人工神经网络算法,建立了碳钢及低合金钢的海水腐蚀预测模型。该模型以合金成分、环境因素为输入参数,以平均腐蚀速率为输出参数。以碳钢、低合金钢的17种钢种在青岛、厦门、榆林海水腐蚀试验站16年腐蚀数据建模。选定A3钢与10CrCuSiV在以上三地16年的腐蚀数据为验证样本。结果表明该网络具有良好的预测精度,能够正确反映海水环境腐蚀性因素及金属材料腐蚀暴露时间与其腐蚀速率的关系,用于碳钢及低合金钢在海洋全浸环境中的腐蚀预测。

Based on the seawater corrosion data collected from China material environmental corrosion test sites, a forecasting model for the seawater corrosion of carbon steel and low alloy steel was built by employing back-propagation neural network （ BP neural network）. In this model, alloy compositions and environment factors were set as input parameters while the average corrosion rate was the output. The sample for the modeling was selected from the seawater corrosion data of 17 kinds of steels, and the data of A3 steel and 10CrCuSiV were used as the verifying samples. All data used were collected from Qingdao, Xiamen, and Yulin seawater test sites with the sampling period of 16 years. The simulation result showed the model has quite good forecast accuracy and it is feasible for the forecast of seawater corrosion.