海水腐蚀速率的不等时距灰色模型与BP神经网络模型组合预测  

Prediction of Marine Corrosion Rate Based on Unequal Interval Grav Model and UP Neural Network

作　　者：经建芳[1] 邓富康 李康春 黄福川[1] 

机构地区：[1]广西大学化学化工学院广西石化资源加工及过程强化技术重点实验室,广西南宁530004

出　　处：《材料保护》2015年第8期33-36,7,共4页Materials Protection

基　　金：南宁市科学研究与技术开发项目(CG20140025)资助

摘　　要：金属在海水中的腐蚀机理及变化规律十分复杂,且采集腐蚀数据存在时间间隔不均匀、数据量小等问题,难以获取准确数据。基于灰色系统理论,提出运用能够适应具有无规律的采集时序数据的不等时距GM(1,1)模型对金属海水腐蚀速率进行建模,并引入了BP人工神经网络模型对预测结果进行残差修正,以提高预测精度。以A3钢与15Mn Mo VN钢腐蚀行为作为实例,进行预测和分析。结果显示:不等时距GM(1,1)与BP神经网络组合预测模型的预测效果明显优于单一预测模型,能更真实地反映海水腐蚀的变化趋势,因而具有较高应用价值。Based on grey system theory,a modeling for prediction of metal corrosion rate in seawater was constructed using an unequal interval gray CM(1,1) model,which could be used for the irregular data sampling time interval acquisition.In addition,the residual error of the model was corrected by the BP neural network to increase the prediction accuracy.Taking the prediction of corrosion behavior of A3 steel and 15 MnMoVN steel as an example,the results showed that the combination model based on unequal interval gray GM(1,1) and BP neural network had higher prediction accuracy than the single model,which reflected the variation tendency of seawater corrosion more exactly,showing a high application value.

关 键 词：海水腐蚀 不等时距GM(1 1)模型 BP神经网络模型 预测 

分 类 号：P755.3[天文地球—海洋科学] TP183[自动化与计算机技术—控制理论与控制工程]