Regression Model for Predicting Expansion of Concrete Exposed to Sulfate Attack Based on Performance-based Classification

10/11/2018 ∙ by Xiangru Jian, et al. ∙ 0

This paper mainly described development of a new kind of regression model to predict long-term expansion of concrete subjected to sulfate-rich environment. The experimental data originally came from a long-term (40+ years), nonaccelerated test program performed by U.S. Bureau of Reclamation (USBR). Expansion data of specimens of 54 different mixtures were measured periodically throughout the whole test. In analysis, the mixtures were firstly classified into three groups using K-means cluster based on different patterns of their expansion performance. Within each group, expansion rate was predicted as an exclusive regression function of water-cement ratio(W/C), C3A content of cement, content of cement or time of expansion. Then Supporting Vector Machine (SVM) helped figure out the criteria of classification relied on characteristic of mixture proportion other than experimental performance, enabling the model to offer prediction for a new mixture without test data. The analysis of the model indicated that concrete with different mixture proportion, especially with different W/C or C3A content, is unlikely to share identical pattern of expansion and should be considered and predicted separately.



There are no comments yet.


This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.