摘要
为有效预测海底管线腐蚀速率,建立了自适应遗传算法(AGA)优化极限学习机(ELM)的预测模型(AGA-ELM)。根据腐蚀影响因素确定输入层参数,腐蚀速率为输出层参数;为避免模型受输入权值矩阵及隐含层偏差随机性的影响,利用AGA对两者进行优化,并采取测试法探究最优隐含层节点数及激活函数。同时,通过改进AGA的遗传算子,提高其寻优性能。以实海挂片试验数据对该模型进行实证分析,对比分析该模型与BP、ELM模型的预测结果。结果表明,AGA-ELM模型泛化性能更好、精准度更高,且预测结果的平均绝对误差、平均绝对百分误差、均方根误差均明显低于其他2种模型,提高了预测的可靠度及吻合度,为海底管线腐蚀预测研究提供了新的研究方向。
In order to effectively predict the corrosion rate of subsea pipelines,a prediction model of adaptive genetic algorithm(AGA)optimized extreme learning machine(ELM)was established.The parameters of the input layer were determined according to the corrosion factors,and the corrosion rate was the parameter of the output layer.In order to avoid the influence of the randomness of input weight matrix and hidden layer deviation,AGA was used to optimize the two parameters,and test method was adopted to explore the optimal number of hidden layer nodes and optimal activation function.At the same time,the performance of AGA was improved by improving the genetic operators.The model was analyzed and compared with the prediction results of BP and ELM models by taking the real sea hanging test.Results showed that AGA-ELM model had better generalization performance and higher precision,and the average absolute error,average absolute percentage error and root mean square error of prediction results were significantly lower than those of the other two models.The reliability and coincidence of the prediction were improved,and a new research direction for the corrosion prediction of subsea pipelines was proposed.
作者
骆正山
李易安
骆济豪
王小完
LUO Zheng-shan;LI Yi-an;LUO Ji-hao;WANG Xiao-wan(School of Management,Xi’an University of Architecture&Technology,Xi’an 710055,China;The High School Affiliated to Xi’an Jiaotong University,Xi’an 710056,China)
出处
《材料保护》
CAS
CSCD
北大核心
2019年第8期51-56,共6页
Materials Protection
基金
国家自然科学基金资助(41877527)
陕西省社科基金项目(2018S34)
