Machine Learning-Based Modeling of Tensile Properties of Glass-Fiber-Reinforced Polymer Pipes under Accelerated Saltwater Aging Conditions

Cristina Roxana Popa¹, Maria Tănase^2,*, Gheorghe Brănoiu³, Elena-Emilia Sirbu^4,5, Cătălina Călin⁴
1 Automatic Control Computers & Electronics Department, Petroleum-Gas University of Ploieşti, Ploiesti, Romania
2 Mechanical Engineering Department, Petroleum-Gas University of Ploieşti, Ploiesti, Romania
3 Petroleum Geology and Reservoir Engineering Department, Petroleum-Gas University of Ploieşti, Ploiesti, Romania
4 Chemistry Department, Petroleum-Gas University of Ploieşti, Ploiesti, Romania
5 National Institute for Research & Development in Chemistry and Petrochemistry ICECHIM, Bucharest, Romania
* Corresponding Author: Maria Tănase. Email: email
(This article belongs to the Special Issue: Computational Modelling of Advanced Polymeric Materials and Structures)

Computer Modeling in Engineering & Sciences https://doi.org/10.32604/cmes.2026.082244

Received 22 March 2026; Accepted 12 May 2026; Published online 29 May 2026

Download PDF

Abstract

Glass-fiber-reinforced polymer (GFRP) pipes are increasingly used in aggressive environments due to their high corrosion resistance and favorable mechanical properties. However, long-term exposure to saline environments and elevated temperatures can lead to degradation of their structural performance. This study investigates the influence of accelerated saltwater aging on the tensile behavior and structural characteristics of GFRP pipes and proposes machine-learning-based predictive models for the ultimate tensile strength (UTS). Experimental specimens were immersed in a 3.5% NaCl solution under controlled temperature and exposure time conditions. Tensile testing revealed that the unexposed samples exhibited a maximum UTS of 79.63 MPa, while aged specimens showed a gradual reduction in strength, although more than 80% of the initial tensile strength was retained after 60 days of exposure. Statistical analysis indicated that temperature was the dominant factor, contributing 60.24% to the variation in UTS, followed by exposure time with 33.72%, with the regression model explaining 93.96% of the total variance (R² = 0.9396). X-ray diffraction analysis revealed a decrease in the degree of crystallinity from 20.49% in the reference sample to 15.05% in the most degraded specimen, corresponding to an approximate 26.5% reduction, which correlated with the observed decline in mechanical strength. Several machine learning approaches were evaluated, including Artificial Neural Networks (ANN), Exponential Gaussian Process Regression, and Boosted Trees. Among them, ANN provided the highest predictive accuracy, demonstrating strong agreement between predicted and experimental UTS values. The results confirm that hydrothermal aging significantly affects both the microstructural and mechanical properties of GFRP pipes, while machine learning models represent effective tools for predicting their long-term performance under aggressive environmental conditions.