10–14 Jun 2025
University of Stavanger
Europe/Oslo timezone

A statistical comparative analysis of fire resistance of reinforced concrete (RC) and concrete-filled steel tubular (CFST) columns using deep learning

Not scheduled
20m
University of Stavanger

University of Stavanger

Oral presentation

Speaker

Afaq Ahmad (Oslo Met)

Description

Fire resistance is a critical aspect which is to be considered to define the structural integrity of columns. This investigation presents the development of predictive neural network (NN) models to predict the fire resistance of reinforced concrete (RC) and concrete filled steel tubular (CFST) columns. Databases of 270 and 223 specimens were prepared for RC and CFST columns, respectively. 12 input parameters were considered for CFST columns, and 8 input parameters were considered for RC columns to train the algorithms. Eight input parameters considered for RC columns modelling are cross-sectional area of columns (mm2), specimen height (mm), compressive strength of concrete (MPa), tensile strength of reinforcement (MPa), longitudinal reinforcement ratio (%), heating rate (°C/min), test load (kN), eccentricity of applied load (mm) and 2 additional parameters were steel plate thickness (mm), and tensile Strength of Steel (MPa) for CFST columns. The models were developed with different neural architectures and most optimum models outperformed all available conventional models with correlation coefficients (R) of 0.92294 for RC columns and 0.99431 for CFST columns. Parametric studies are also done to observe and compare the effectiveness of the most important parameters on fire resistance of RC and CFST columns. The developed models encompass significant advantages for the field of fire engineering. These models can be used to evaluate and compare the performance of RC and CFST columns in the event of fire.

Primary authors

Afaq Ahmad (Oslo Met) Muhammad Noman (University of Florida) Muhammad Faizan (UET Taxila) Muhammad Salman (UET Taxila)

Presentation materials

There are no materials yet.

Peer reviewing

Paper

Paper files: