Practical Machine Learning Application for Predicting Axial Capacity of Composite Concrete-Filled Steel Tube Columns Considering Effect of Cross-Sectional Shapes

Previous studies on machine-learning (ML) prediction of axial capacity of composite concrete-filled steel tubular (CFST) columns under axial loading relate mainly to only one cross-section shape, meaning that they are limited to only one given application. In this paper, a ML model—namely support ve...

Full description

Saved in:
Bibliographic Details
Published in:International journal of steel structures 2023, 23(1), , pp.263-278
Main Authors: Duong, T. H., Le, T.-T., Le, M. V.
Format: Article
Language:English
Subjects:
Axial loads
Civil Engineering
Concrete columns
Cross-sections
Engineering
Machine learning
Materials Science
Sensitivity analysis
Solid Mechanics
Steel columns
Steel tubes
Support vector machines
토목공학
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Previous studies on machine-learning (ML) prediction of axial capacity of composite concrete-filled steel tubular (CFST) columns under axial loading relate mainly to only one cross-section shape, meaning that they are limited to only one given application. In this paper, a ML model—namely support vector machine (SVM)—is proposed for the prediction of CFST columns with different cross-section shapes: circular, elliptical, square and rectangular, because they are the most widely used in engineering structures. A database consisting of 1093 tests was gathered from the available literature and used to train and validate the SVM model. The model’s performance was quantified by various performance indicators: coefficient of determination (R 2 ), root mean squared error, mean absolute error, Willmott’s index of agreement, and mean absolute percentage error. Based on the SVM model, sensitivity analysis, influence of different factors, parametric study and comparison with the literature are presented. A graphic user interface of the proposed model was also implemented. The model could be extended to study the effect of other cross-sectional shapes, making for wider applicability.
ISSN:1598-2351
2093-6311
DOI:10.1007/s13296-022-00693-0