SigmaBuckling

Overview

Predicting impact-induced stress in steel columns using ML and section-based modeling

SigmaBuckling is a machine learning-powered platform for predicting the structural response of multi-cell steel columns subjected to axial impact. Built upon a high-fidelity dataset generated via finite element simulations, this software estimates the peak impact force and stress development within complex cross-sectional geometries in real time.

Developed based on a peer-reviewed scientific model published in the Journal of Constructional Steel Research (Elsevier), SigmaBuckling ensures the scientific credibility required for academic research, industrial practice, and design optimization workflows.

Academic Publication:
Y. Yılmaz, F. Öztürk, S. Demir, and A. D. Demir, "Prediction of load-bearing capacity of sigma section CFS beam-column members using ensemble and deep learning algorithms," Journal of Constructional Steel Research, vol. 228, p. 109458, May 2025, doi:10.1016/j.jcsr.2025.109458.

Core Features

• Axial Impact Stress Prediction: Real-time forecast of σₘₐₓ under axial dynamic loading.
• Section Geometry Integration: Cross-sectional shape, thickness, and internal cell configurations fully considered.
• Velocity-Sensitive Modeling: Captures dynamic strain rate effects in high-speed impacts.
• High Model Accuracy: Gradient Boosting algorithm with R² 0.97 and MAE of 10.8 kN.
• FEA-Free Simulation: No need for ABAQUS or LS-DYNA. Pure ML-based inference in <1 second.

Input Parameters

Output Parameters

• Peak axial force (kN)
• Maximum shortening (mm)
• Stress distribution factor
• Downloadable reports (CSV + PDF)
• Stress–velocity visual chart (PNG)

Target Users

• Structural engineers in crash design
• Impact mechanics researchers
• Automotive and logistics infrastructure planners
• Advanced civil engineering analysts

Disclaimer

This software predicts impact-related structural behavior of multi-cell steel columns based on experimental and numerical data reflecting collapse-stage loading. The estimated outputs correspond to peak responses near or at the failure limit.

Therefore, any capacity reductions, partial safety factors, or design-specific modifications must be applied by the responsible structural engineer in accordance with national codes and engineering judgment. The final interpretation and use of results are solely the responsibility of the user.

The underlying machine learning model has been developed from the peer-reviewed study published in a high-impact academic journal:
Y. Yılmaz, F. Öztürk, S. Demir, A. Durmuş Demir (2025). Prediction of load-bearing capacity of sigma section CFS beam-column members using ensemble and deep learning algorithms. Journal of Constructional Steel Research, 228, 109458. https://doi.org/10.1016/j.jcsr.2025.109458

This publication in a scientifically reputable journal ensures the academic credibility of the software, but it does not replace professional responsibility or regulatory compliance in engineering design.