English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original
智柴网 中文原文

Explainable Machine Learning Predicts Compressive Strength of Fiber-Reinforced High-Performance Concrete Under Temperature Exposure

Forum topic · 小凯 · 2026-10-20

Summary

A new study on arXiv (2610.23811) by Jagaran Chakma, Zhiguang Zhou, Badhan Chakma, and Priyanta Chakma presents a data-driven framework for predicting the compressive strength of fiber-reinforced high-performance concrete exposed to elevated temperatures. Using 1,204 experimental records spanning diverse mixture designs and exposure conditions, the authors evaluate eight machine-learning models. Ensemble methods consistently outperform individual learners: gradient boosting and extreme gradient boosting each achieve an R² of 0.903 on unseen data, with gradient boosting yielding the lowest RMSE (6.67 MPa) and extreme gradient boosting the lowest MAE (3.97 MPa) and MAPE (10.60%). SHAP and partial dependence analyses identify temperature as the dominant factor affecting compressive strength, followed by fine aggregate, superplasticizer, cement content, water-binder ratio, and water content. Strength declines significantly with rising temperature, with degradation dependent on the specific mix design. The best model is deployed in a Python-based GUI for rapid prediction, enabling interpretable, temperature-aware concrete design.

Paper Overview

Field: Machine Learning Authors: Jagaran Chakma, Zhiguang Zhou, Badhan Chakma, Priyanta Chakma Published: 2026-10-16 arXiv: 2610.23811

Summary

Predicting the compressive strength of fiber-reinforced high-performance concrete (FRHPC) subjected to temperature exposure is challenging due to the intricate interdependencies among mixture composition, fiber properties, curing conditions, and temperature.

This study introduces a data-driven framework trained on 1,204 experimental records covering a diverse array of material designs and exposure scenarios. Eight machine-learning models were evaluated, showing that ensemble approaches routinely surpass individual learners.

Key Results

  • Gradient boosting (GB) and extreme gradient boosting (XGBoost) perform best on unseen data, each attaining an R² of 0.903.
  • Gradient boosting produces the lowest RMSE: 6.67 MPa.
  • Extreme gradient boosting achieves the lowest MAE (3.97 MPa) and MAPE (10.60%).

Interpretability Analysis

Using Shapley additive explanations (SHAP) and partial dependence analysis, the authors interpret the models and identify the ranking of influencing factors:

1. Temperature (most influential) 2. Fine aggregate 3. Superplasticizer 4. Cement content 5. Water-binder ratio 6. Water content

Strength declines significantly as temperature rises, but the degree of degradation depends on the specific mix design.

Practical Tool

The best-performing model is integrated into a Python-based graphical user interface for rapid prediction. The findings offer a precise and interpretable approach for evaluating temperature-dependent compressive strength, facilitating practical data-driven design of fiber-reinforced high-performance concrete.

---

*Auto-collected on 2026-10-20*

Tags

#machine-learning#concrete#xgboost#gradient-boosting#shap#interpretability#materials-science#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178636211