基于最小二乘支持向量机的桥墩局部冲刷深度预测方法  

Prediction Method of Local Scour Depth of Bridge Pier Based on Least Squares Support Vector Machines

作　　者：王秋生[1] 邓洪森 周鹏展 WANG Qiu-sheng;DENG Hong-sen;ZHOU Peng-zhan(Key Laboratory of Urban Security and Disaster Engineering of Ministry of Education,Beijing University of Technology,Beijing 100124,China)

机构地区：[1]北京工业大学城市与工程安全减灾教育部重点实验室,北京100124

出　　处：《水电能源科学》2022年第1期120-123,共4页Water Resources and Power

基　　金：国家自然科学基金项目(51679003);北京市科学技术委员会重大专项基金(Z161100002216001)。

摘　　要：桥墩局部冲刷是引起桥梁破坏的重要原因,采用现行规范和经验公式所得计算值偏于保守且过于离散。对此,选取美国交通运输部的实测数据中代表冲刷条件的6个参数,应用主成分分析法处理原始数据,提取出3个主成分,消除了物理参数之间的多重共线性。运用最小二乘支持向量机方法(LS-SVM)分别对原始数据和主成分进行拟合,并通过均方误差(M;)和决定系数(R;)两个统计参数评判拟合效果。结果表明,两种输入数据预测结果均优于现行规范计算结果;采用主成分进行计算要优于采用原始参数拟合的结果,采用主成分进行拟合的决定系数高达0.97,最低为0.75。Local scour of bridge pier is an important cause of bridge failure.The calculated values based on existing codes and empirical formulas are conservative and too discrete.In this paper,six parameters representing scour conditions were selected from the measured data of the US Department of Transportation,and three principal components were extracted from the original data by principal component analysis method,eliminating the multicollinearity between physical parameters.The least square support vector machine(LS-SVM)was used to fit the original data and principal components respectively,and the fitting effect was evaluated by two statistical parameters:mean square error(M;)and determinant coefficient(R;).The results show that the prediction results of the two kinds of input data are better than those calculated by the current specification.The result of principal component fitting is better than that of original parameter fitting.The determination coefficient of principal component fitting is as high as 0.97 and the lowest is 0.75.

关 键 词：桥墩冲刷 机器学习 LS-SVM 主成分分析 

分 类 号：TV698.1[水利工程—水利水电工程]