Explainable machine learning prediction of the flexural capacity of UHPFRC beams

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21610%2F25%3A00385806" target="_blank" >RIV/68407700:21610/25:00385806 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.1201/9781003677895-130" target="_blank" >https://doi.org/10.1201/9781003677895-130</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1201/9781003677895-130" target="_blank" >10.1201/9781003677895-130</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Explainable machine learning prediction of the flexural capacity of UHPFRC beams

Popis výsledku v původním jazyce

This study developed explainable machine learning models to provide innovative and accurate predictions of the flexural capacity of ultra-high-performance fibre-reinforced concrete (UHPFRC) beams. The study used a data-driven approach that involved compiling the dataset from the existing experimental tests on the flexural behaviour of UHPFRC beams and training machine learning models. Key input variables included; the compressive strength of UHPFRC, tensile strength of longitudinal rebar, yield strength of steel, cross-section dimensions, and fibre characteristics. Machine learning models that were employed include; Artificial Neural Network (ANN), K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), Gradient Boosting Machine (GBM), CatBoost, and Extreme Gradient Boosting (XGBoost). The training involved splitting the dataset into 80% training and 20% testing with 10-fold cross-validation and grid search used in hyperparameter tuning. R-squared, Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) were used to evaluate the performance of the models and a technique known as Shapley additive explanations (SHAP) was used to interpret the models. It was found that ensemble machine learning models, particularly XGBoost, outperform individual ML models and traditional code equations, suggesting the potential of machine learning to improve design accuracy. The study also revealed that geometric dimensions and rebar characteristics are the most influential input variables affecting the flexural capacity of UHPFRC beams, indicating that careful consideration of these factors can lead to effective and optimal designs.

Název v anglickém jazyce

Explainable machine learning prediction of the flexural capacity of UHPFRC beams

Popis výsledku anglicky

This study developed explainable machine learning models to provide innovative and accurate predictions of the flexural capacity of ultra-high-performance fibre-reinforced concrete (UHPFRC) beams. The study used a data-driven approach that involved compiling the dataset from the existing experimental tests on the flexural behaviour of UHPFRC beams and training machine learning models. Key input variables included; the compressive strength of UHPFRC, tensile strength of longitudinal rebar, yield strength of steel, cross-section dimensions, and fibre characteristics. Machine learning models that were employed include; Artificial Neural Network (ANN), K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), Gradient Boosting Machine (GBM), CatBoost, and Extreme Gradient Boosting (XGBoost). The training involved splitting the dataset into 80% training and 20% testing with 10-fold cross-validation and grid search used in hyperparameter tuning. R-squared, Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) were used to evaluate the performance of the models and a technique known as Shapley additive explanations (SHAP) was used to interpret the models. It was found that ensemble machine learning models, particularly XGBoost, outperform individual ML models and traditional code equations, suggesting the potential of machine learning to improve design accuracy. The study also revealed that geometric dimensions and rebar characteristics are the most influential input variables affecting the flexural capacity of UHPFRC beams, indicating that careful consideration of these factors can lead to effective and optimal designs.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

20102 - Construction engineering, Municipal and structural engineering

Projekt

—

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2025

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Engineering Materials, Structures, Systems and Methods for a More Sustainable Future

ISBN

978-1-003-67789-5

ISSN

—

e-ISSN

—

Počet stran výsledku

5

Strana od-do

—

Název nakladatele

TAYLOR & FRANCIS LTD

Místo vydání

ABINGDON, OXON

Místo konání akce

Cape Town

Datum konání akce

1. 9. 2025

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—