Prediction of Moment Capacity of Ultra-High-Performance Concrete Beams Using Explainable Extreme Gradient Boosting Machine Learning Model

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21610%2F24%3A00377294" target="_blank" >RIV/68407700:21610/24:00377294 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Prediction of Moment Capacity of Ultra-High-Performance Concrete Beams Using Explainable Extreme Gradient Boosting Machine Learning Model

Popis výsledku v původním jazyce

The utilisation of Ultra-High-Performance Concrete (UHPC) in structural engineering presents significant design challenges, often necessitating the use of costly and complex finite element analyses due to strongly non-linear material behaviour. To address these issues, this study introduces an explainable machine learning (ML) model employing the eXtreme Gradient Boosting (XGBoost) algorithm, aimed at accurately predicting the bending capacity of UHPC beams. Demonstrating robust performance, the XGBoost model achieved a high coefficient of determination (R2) of 99.5% (training) and 97.6% (testing), outperforming conventional analytical models (FHWA, SIA 2052,and JGJ/T 465) cited in the literature. Further analysis employing the SHapley Additive Explanation (SHAP) framework provided insights into the model’s decision-making process, revealing that the effective depth, steel reinforcement ratio, and steel yield strength significantly influence the predicted bending capacity, while the aspect ratio of fibres has a minimal impact. This research highlights the potential of the XGBoost model to revolutionise prediction, analysis, and design processes within performance-based design frameworks, offering a promising alternative to traditional non-linear analysis methods for UHPC beams.

Název v anglickém jazyce

Prediction of Moment Capacity of Ultra-High-Performance Concrete Beams Using Explainable Extreme Gradient Boosting Machine Learning Model

Popis výsledku anglicky

The utilisation of Ultra-High-Performance Concrete (UHPC) in structural engineering presents significant design challenges, often necessitating the use of costly and complex finite element analyses due to strongly non-linear material behaviour. To address these issues, this study introduces an explainable machine learning (ML) model employing the eXtreme Gradient Boosting (XGBoost) algorithm, aimed at accurately predicting the bending capacity of UHPC beams. Demonstrating robust performance, the XGBoost model achieved a high coefficient of determination (R2) of 99.5% (training) and 97.6% (testing), outperforming conventional analytical models (FHWA, SIA 2052,and JGJ/T 465) cited in the literature. Further analysis employing the SHapley Additive Explanation (SHAP) framework provided insights into the model’s decision-making process, revealing that the effective depth, steel reinforcement ratio, and steel yield strength significantly influence the predicted bending capacity, while the aspect ratio of fibres has a minimal impact. This research highlights the potential of the XGBoost model to revolutionise prediction, analysis, and design processes within performance-based design frameworks, offering a promising alternative to traditional non-linear analysis methods for UHPC beams.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

20102 - Construction engineering, Municipal and structural engineering

Projekt

<a href="/cs/project/GA24-10892S" target="_blank" >GA24-10892S: Strojové učení pro víceúrovňové modelování prostorové variability a trhlin pro zajištění udržitelnosti betonových konstrukcí</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2024

Kód důvěrnosti údajů

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

Název statě ve sborníku

1st International Conference on Synergy between Multi-physics/Multi-scale Modeling and Machine Learning

ISBN

978-9926-8888-1-7

ISSN

—

e-ISSN

—

Počet stran výsledku

4

Strana od-do

84-87

Název nakladatele

Association of Computational Mechanics in Bosnia and Herzegovina

Místo vydání

Sarajevo

Místo konání akce

Prague

Datum konání akce

19. 6. 2024

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—