摘要
建立人工神经网络 ,并利用 4个实海暴露站点 8年累积的腐蚀数据对其进行训练 .在训练成功以后 ,应用其对合金的 1 6年腐蚀行为进行预测 .预测结果与实际结果的误差在 2 0 %以内 ,远比传统的函数回归方法小 .尤其对于规律性差 ,无法成功进行函数回归的腐蚀数据 。
Set up an artificial neural net and train it with the data provided by four seawater corrosion experiment stations.After the succession of training,the net can predict any kind of metals corrosion status on 16 years and the comparative error is not beyond 20 percent.Compare with the way of function regression,this result is more accurate.The net can also trained with the anomaly data that can not regress to any function and the the net can make an accurate prediction.
出处
《中国腐蚀与防护学报》
CAS
CSCD
2004年第1期29-32,共4页
Journal of Chinese Society For Corrosion and Protection
基金
国家自然科学基金"我国海域海水对铝合金的腐蚀性研究"资助 ( 5 0 0 710 12 )
关键词
人工神经网络
海水腐蚀
腐蚀预测
artificial neural net,seawater corrosion,the prediction of corrosion
