基于随机森林和支持向量机的高性能混凝土抗渗性预测研究

为了实现对混凝土抗渗性快速而精确地预测,提出了一种基于随机森林(RF)和支持向量机(SVM)的RF-SVM预测模型。首先以氯离子渗透系数为抗渗性评价指标,基于原材料配比确定了混凝土抗渗性的初始指标体系,然后利用随机森林算法结合后向剔除法进行指标筛选,剔除了冗余指标,得到了用于支持向量机建模的最优指标集,最后在此基础上建立了基于支持向量机的混凝土抗渗性预测模型,并研发了RF-SVM算法。以东北某高速公路项目为背景进行应用分析,结果表明,所提出的RF-SVM模型能够有效筛除冗余因素,得到精度较高的预测结果,且预测结果满足工程实践的要求,能够为混凝土抗渗性预测提供一种快速有效的方法。

Prediction of Impermeability of Concrete Based on Random Forest and Support Vector Machine

In order to predict the impermeability of concrete quickly and accurately, a RF-SVM prediction model based on random forest (RF) and support vector machine (SVM) was proposed. At first, the permeability coefficient of chloride ion was taken as the evaluation index of impermeability, the initial index system of concrete impermeability was determined based on the ratio of raw materials.Then, the random forest algorithm combined with backward elimination method was used to screen the indexes, and the redundant indexes were eliminated, and the optimal set of indexes for support vector machine modeling was obtained. Finally, a prediction model of concrete impermeability based on support vector machine was established, and a RF-SVM algorithm was developed. Based on a highway project in northeast China, the results show that the proposed RF-SVM model effectively screens out the redundant factors and obtains high precision prediction results, which meets the requirements of engineering practice, and provides a fast and effective method for predicting the impermeability of concrete.