Hybrid intelligent model for approximating unconfined compressive strength of cement-based bricks with odd-valued array of peat content (0-29%)

Title
Hybrid intelligent model for approximating unconfined compressive strength of cement-based bricks with odd-valued array of peat content (0-29%)
Authors
송기일
Keywords
Forecasting, Construction, Peat, Estimation, Soft computing, Hybrid
Issue Date
2015-11
Publisher
POWDER TECHNOLOGY
Series/Report no.
POWDER TECHNOLOGY ; Vol284 Startpage 560 Endpage 570
Abstract
This article presents an innovative approach to estimate the unconfined compressive strength (UCS) of peat-enhanced bricks using a hybrid intelligent system (HIS) resulting from integration of support vector regression (SVR) and Bat meta-heuristic algorithm (hereafter, Bat–SVR). First, peat-enhanced brick specimens were prepared for various compositions of cement, sand, and peat (odd-valued array of peat inclusion in the range of 0–29% from the total specimens' weight). Further, the experimental works were carried out to obtain the UCS of specimens in different curing period. Finally, HIS model was used to predict the UCS of cement–peat–soil mixture. Basically, we used a newly-developed Bat algorithm for tuning the SVR parameters, because the accuracy of SVR estimation highly relies on these parameters. Results from the experimental study were used to train and estimate the UCS of peat-enhanced bricks. In addition, we compared the accuracy of the developed HIS model to other conventional soft computing techniques (i.e., ANFIS and neural network). It was found that the proposed approach outperforms the other conventional prediction models and better estimates the UCS of peat-enhanced bricks.
URI
https://www.sciencedirect.com/science/article/pii/S0032591015005586
http://dspace.inha.ac.kr/handle/10505/55246
ISSN
0032-5910
Appears in Collections:
College of Engineering(공과대학) > Civil Engineering (사회인프라공학) > Journal Papers, Reports(사회인프라공학 논문, 보고서)
Files in This Item:
1152_Hybrid intelligent model.pdfDownload

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Browse