StiffPredict

Overview

Rotational stiffness prediction of rack beam-to-upright connections using machine learning.

StiffPredict is an advanced machine learning-based tool for predicting the rotational stiffness (k₀) of semi-rigid beam-to-upright connections in storage rack systems. Developed from a validated dataset obtained through experimental investigations and numerical simulations, StiffPredict provides reliable estimations of joint stiffness based on connector geometry, material properties, and structural configuration.

The underlying model reflects state-of-the-art research on semi-rigid connections, ensuring transparency, credibility, and engineering-grade accuracy for structural engineers and researchers engaged in the design and analysis of cold-formed steel storage racks.

Academic Publication:
Yazici, C., & Dominguez-Gutierrez, F. J. (2026). Data-driven prediction of initial rotational stiffness in beam-to-upright connections of steel pallet racks. Challenge Journal of Structural Mechanics, 12(2), 73-86. https://doi.org/10.20528/cjsmec.2026.02.002

Core Features

• Rotational stiffness (k₀) prediction for semi-rigid beam-to-upright joints
• Integration of connector geometry, upright properties, and material characteristics
• Automated multi-model machine learning framework with accuracy optimization
• Real-time estimation without the need for full-scale testing
• Model explainability through feature importance analysis
• Exportable performance reports and graphical visualization tools

Inputs

Outputs

• Predicted rotational stiffness (k₀, kNm/rad)
• Accuracy indicators (R², RMSE, MAE)
• Normalized feature contributions
• Exportable technical report (PDF/CSV)
• Visual moment–rotation charts

Target Users

• Rack system designers and practicing engineers
• Structural consultants and design offices
• Researchers in cold-formed steel and semi-rigid connection mechanics
• Warehouse safety and infrastructure specialists

Disclaimer

This software predicts the rotational stiffness of beam-to-upright connections under idealized loading conditions. To ensure conservative design, users are advised to apply a 20% reduction to the predicted k₀ values.

All design decisions, safety factors, and regulatory verifications must be independently performed by the practicing engineer. The developers assume no liability for misuse or misinterpretation of results.

The underlying methodology has been documented and published in a peer-reviewed academic paper. The publication details are provided in the Academic Publication section.