PSI - Issue 70

N. Vignesh Kumar et al. / Procedia Structural Integrity 70 (2025) 453–460

454

1. Introduction Fiber Reinforced Polymer (FRP) has marked its place as a confinement material for concrete structures thereby enhancing its mechanical properties (Shang et al., 2025). FRP confined concrete elements are restrained to lateral expansion under axial loading, thereby enhancing the compressive strength of the concrete (Amin Mozumder et al., 2016). FRP confinement method has been acknowledged widely due to its high tensile strength to weight ratio, less susceptible to environmental factors, less laborious during application and least influence on structural dimensions (Valasaki & G.Papakonstantinou, 2023). Ensuring the prediction of the compressive strength of FRP confined concrete is crucial for optimal design in the usage of FRP as well as assessing the performance of FRP confined concrete elements (Napoli & Realfonzo, 2020). Over the years, many analytical and empirical models have been devised to predict the FRP confined concrete strength based on the properties of the concrete and the confinement material (Shayanfar et al., 2021). However, those models lagged in capturing the complex, non-linear relationships existed between the variables used in predicting the strength of FRP confined concrete (llyas et al., 2022). This research article aims to evaluate and compare the performance of 5 machine learning techniques – Stepwise regression, LASSO regression, Support Vector Machine (SVM), Regression Tree Ensemble (RTE) and Gaussian Process Regression (GPR) for predicting the strength of FRP confined concrete cylinders. The objectives of this study include: (1) Developing prediction models using a preprocessed database derived from Realfonzo and Napoli (2011); (2) Comparing the performance of these models using established metrics; (3) Identifying the most influential parameters using sensitivity analysis; and (4) Providing insights into the relative advantages of different machine learning techniques for this strength prediction of FRP confined concrete cylinders. The article is structured as follows: Section 2 shows a brief review of relevant literature; Section 3 provides the methodology, database and dataset characteristics, preprocessing approach and deployment of machine learning ML) techniques; Section 4 presents the results and discussion; and Section 5 provides the conclusion and the scope for future research work.

Nomenclature f co

Unconfined compressive strength of concrete Peak compressive strength of FRP confined concrete Tensile strength of Fiber Reinforce Polymer (FRP) material

f cc f frp t frp

Thickness of FRP jacket

Diameter of the concrete cylinder specimen Height to diameter ratio of the cylinder specimen

D

H_D

Strain efficiency factor

k e R 2

Coefficient of determination, indicating model fit RMSE Root Mean Square Error, measuring prediction accuracy MAPE Mean Absolute Percentage Error, indicating percentage deviation in predictions

2 Literature Review Over the past few years, the nature of FRP confined concrete has been analyzed and that paved to numerous analytical and empirical models for the prediction of FRP confined concrete strength (Nain et al., 2020). Generally, these models can be classified into two categories such as Analysis – Oriented models which are based on iteration of numerical procedures comprising the physical variable interaction pertaining to concrete and FRP at each experimental loading stage, and Design – Oriented models provide comprehendible equations for direct strength prediction of FRP confined concrete (Gharaei-Moghaddam et al., 2023). One among the design-oriented models has wide acceptance and has been considered into design guidelines. Their approach related the unconfined concrete strength and the FRP confined concrete strength and the lateral confining pressure provided by the FRP confinement. The refined approach subsequently incorporated the effects of FRP stiffness and concrete strength on the effectiveness of confinement.