Using Support Vector Machine and Logistic Regression Methods to Build Groutability Models for Permeation Grouting with Microfine Cement Grout
Date Issued
2012
Date
2012
Author(s)
Lai, Po-Chou
Abstract
The purpose of this research is to establish the prediction model of the groutability of the silty sand soils using microfine cement grouts in a permeation grouting. Due to the fact that the region covered in this paper consists of the silty sand soils with relatively higher proportion of the fines content(FC) and the particle size of microfine cement used is considerably smaller than the conventional Portland cement, the existing empirical formula with relative particle size ratio is unable to provide effective predictions. Thus, this research derives the prediction model and formula from 240 data in Taiwan (Taipei and Kaohsiung) using Support Vector Machine(SVM) with Tabu Search(TS) and Logistic Regression(LR), respectively. In terms of selecting factors for the groutability, apart from the relative size for particles passing through soil with 10% and 15% permeability that are used in the conventional empirical formula with relative particle size ratio, this research also takes the fines content(FC) and the water-to-cement ratio(W/C) into account. By using SVM with TS, the model established can reach 97.75% precision of prediction. Moreover, the fine results of groutability prediction, not only indicate the feasibility of applying SVM with TS, but also explain the advantages of SVM in dealing with complicated and non-linear scenarios. In addition, the prediction formula derived from LR shares the same simplicity as in the conventional empirical formula with relative particle size ratio. It is hoped that, since engineers can use this formula with ease, it can also be widely used in applications and real-life constructions.
Subjects
Support Vector Machine(SVM)
Tabu Search algorithm (TS)
Logistic Regression(LR)
microfine cement
permeation grouting
groutability
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-101-R99622012-1.pdf
Size
23.54 KB
Format
Adobe PDF
Checksum
(MD5):c4f5054979fa3c366b3cfd1a853d2ffa