Open AccessPrediction of impermeability of the concrete structure based on random forest and support vector machine
Author(s) -
Lei Wang,
Xianguo Wu,
Hongyu Chen,
Tiemei Zeng
Publication year - 2020
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/552/1/012004
Subject(s) - support vector machine , random forest , durability , aggregate (composite) , service life , predictive modelling , computer science , artificial neural network , index (typography) , permeability (electromagnetism) , machine learning , environmental science , artificial intelligence , engineering , materials science , reliability engineering , database , composite material , chemistry , biochemistry , membrane , world wide web
The durability of concrete has a significant impact on the service life. Impermeability is one of the important indicators of concrete durability. It is of great significance to quickly and reasonably predict the impermeability of concrete. This paper combines random forest and support vector machine (RF-SVM) methods. Taking a highway project as the research background, 11 factors were selected as the impact index of concrete impermeability, and the chloride permeability coefficient was used as the evaluation index of concrete impermeability. After random forest index screening, six factors including water-binder ratio, cement dosage, cement strength, fine aggregate, water-reducing agent and coarse aggregate were selected to construct a support vector machine model to predict the impermeability of concrete. The prediction results of the RF-SVM model are compared with the BP neural network model and the support vector machine model without index screening. The results show that the RF-SVM model has higher prediction accuracy and better fitting effect, which provides an effective method for the prediction of concrete impermeability.