基于支持向量机回归的粉煤灰混凝土氯离子质量分数预测  

Prediction of Chloride Ions Mass Fraction in Fly Ash Concrete based on Support Vector Regression

作　　者：王龙龙 余威龙 章玉容 WANG Longlong;YU Weilong;ZHANG Yurong

机构地区：[1]杭州市商贸旅游集团有限公司,浙江杭州310003 [2]华信咨询设计研究院有限公司,浙江杭州310052 [3]浙江工业大学土木工程学院,浙江杭州310014

出　　处：《浙江建筑》2024年第3期79-83,共5页Zhejiang Construction

摘　　要：基于自然潮差环境下粉煤灰混凝土长期暴露试验,获取了3150组自由氯离子质量分数数据,建立了基于支持向量机回归方法(Support Vector Regression,SVR)的粉煤灰混凝土中自由氯离子质量分数预测模型。该模型研究了数据预处理方法,核函数以及超参数优化方法对自由氯离子质量分数预测精度的影响,分析了水灰比、粉煤灰掺量、暴露时间和渗透深度4个输入参数对自由氯离子质量分数预测结果的影响大小。同时,开展基于未测参数的自由氯离子质量分数预测。结果表明:当采用归一化的数据预处理方式,并使用径向基核函数及贝叶斯优化算法时,自由氯离子质量分数预测结果最佳。当自由氯离子质量分数小于0.1%时,所构建的SVR模型得到的预测值与实际氯离子质量分数存在较大差距。Based on a long-term exposure test under natural tidal environment,3,150 groups of data on free chloride ion mass fraction were obtained and used to established the chloride ion mass fraction prediction model by the support vector regression(SVR)method.We proposed a prediction model for the chloride ion mass fraction in fly ash concrete.The established SVR model analyzed the influences of data preprocessing,kernel function and hyperparameter optimization on the accuracy of prediction results.Besides,the effects of four input parameters,including water-cement ratio,fly ash content,exposure time and permeation depth,on the prediction results were explored.Meanwhile,predictions of free chloride ion mass fraction based on unmeasured parameters were also conducted.Results showed that the best prediction results for free chloride ion mass fraction were obtained by using a normalized data preprocessing method,RBF kerne function and Bayesian optimization hyperparameter optimization method.When the free chloride ion mass fraction was less than 0.1%,there were an obvious difference between prediction values obtained by the SVR method and the measured free chloride ion mass fraction.

关 键 词：自由氯离子质量分数 支持向量机回归 粉煤灰混凝土 预测 

分 类 号：TU528[建筑科学—建筑技术科学]