An Automated AI-Based Framework for Rapid Progressive Collapse Assessment of Steel Frames

Progressive collapse remains one of the most critical failure mechanisms in multi-story steel structures, whereby the loss of a single structural element can trigger disproportionate failure of the surrounding system. Although international design guidelines increasingly require explicit collapse-resistance evaluation, practical implementation remains limited due to the high computational cost and modelling complexity of nonlinear finite element (FE) analysis. This study presents an automated artificial intelligence-based framework for rapid progressive collapse assessment of multi-story steel moment frames, jointly predicting the Load Increase Factor (LIF) and Dynamic Increase Factor (DIF) without repeated nonlinear FE analysis. The specific contribution of this work is threefold: (1) a large-scale parametric database generated through a factorial study spanning five geometric and configurational parameters, number of stories, story height, bay span, number of bays, and column removal location, producing 10,962 unique base building configurations, structured as 32,886 configuration–scenario records, each jointly labelled with LIF and DIF; (2) a systematic benchmarking of seven machine learning algorithms and four deep neural network architectures, evaluated on a configuration-level data split in which all load scenarios of a configuration remain in the same partition, and selected through an overfitting-aware criterion rather than validation accuracy alone; and (3) the identification and validation of two accurate surrogate models, a gradient-boosting model (LightGBM) and a compact deep neural network with three hidden layers of decreasing width and rectified linear unit activations, that predict LIF and DIF directly from frame geometry and loading. The deep neural network (DL-3 Deep) achieved the highest test-set accuracy, with coefficients of determination of 0.9714 (LIF) and 0.9576 (DIF) and mean absolute percentage errors of 2.90% and 3.51%, while LightGBM achieved comparable accuracy (0.9664 and 0.9535) as the best classical benchmark. The deep neural network was accordingly deployed as a publicly accessible, web-based decision-support platform that returns predictions in a fraction of a second. Comparison with six held-out FE configurations spanning the validated domain gave mean absolute deviations of 3.76% (LIF) and 3.74% (DIF) for the deep neural network. These findings highlight the feasibility of data-driven surrogate modelling as a practical, deployable surrogate for rapid preliminary progressive-collapse screening within the validated parameter domain, rather than a general replacement for simulation-intensive assessment.

Authors

Institutions

Publication Details

Journal
Civil and Environmental Engineering
Published
2026-11-18
DOI
https://doi.org/10.2478/cee-2027-0030
Primary Topic
Structural Response to Dynamic Loads
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

An Automated AI-Based Framework for Rapid Progressive Collapse Assessment of Steel Frames

AbdElMoniem Bayoumi, Sherif A. Mourad, Amr Mansour, Mohamed H. Serror
3 links
Civil and Environmental Engineering
Structural Response to Dynamic Loads
article

An Automated AI-Based Framework for Rapid Progressive Collapse Assessment of Steel Frames

AbdElMoniem Bayoumi, Sherif A. Mourad, Amr Mansour, Mohamed H. Serror
article en

Abstract

Progressive collapse remains one of the most critical failure mechanisms in multi-story steel structures, whereby the loss of a single structural element can trigger disproportionate failure of the surrounding system. Although international design guidelines increasingly require explicit collapse-resistance evaluation, practical implementation remains limited due to the high computational cost and modelling complexity of nonlinear finite element (FE) analysis. This study presents an automated artificial intelligence-based framework for rapid progressive collapse assessment of multi-story steel moment frames, jointly predicting the Load Increase Factor (LIF) and Dynamic Increase Factor (DIF) without repeated nonlinear FE analysis. The specific contribution of this work is threefold: (1) a large-scale parametric database generated through a factorial study spanning five geometric and configurational parameters, number of stories, story height, bay span, number of bays, and column removal location, producing 10,962 unique base building configurations, structured as 32,886 configuration–scenario records, each jointly labelled with LIF and DIF; (2) a systematic benchmarking of seven machine learning algorithms and four deep neural network architectures, evaluated on a configuration-level data split in which all load scenarios of a configuration remain in the same partition, and selected through an overfitting-aware criterion rather than validation accuracy alone; and (3) the identification and validation of two accurate surrogate models, a gradient-boosting model (LightGBM) and a compact deep neural network with three hidden layers of decreasing width and rectified linear unit activations, that predict LIF and DIF directly from frame geometry and loading. The deep neural network (DL-3 Deep) achieved the highest test-set accuracy, with coefficients of determination of 0.9714 (LIF) and 0.9576 (DIF) and mean absolute percentage errors of 2.90% and 3.51%, while LightGBM achieved comparable accuracy (0.9664 and 0.9535) as the best classical benchmark. The deep neural network was accordingly deployed as a publicly accessible, web-based decision-support platform that returns predictions in a fraction of a second. Comparison with six held-out FE configurations spanning the validated domain gave mean absolute deviations of 3.76% (LIF) and 3.74% (DIF) for the deep neural network. These findings highlight the feasibility of data-driven surrogate modelling as a practical, deployable surrogate for rapid preliminary progressive-collapse screening within the validated parameter domain, rather than a general replacement for simulation-intensive assessment.

Civil and Environmental Engineering
Cairo University (EG)
Industry, innovation and infrastructure
Structural Response to Dynamic Loads
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon