TY - JOUR
T1 - Using machine learning to predict the long-term performance of fibre-reinforced polymer structures
T2 - A state-of-the-art review
AU - Machello, Chiara
AU - Bazli, Milad
AU - Rajabipour, Ali
AU - Mehdizadeh Rad, Hooman
AU - Arashpour, Mehrdad
AU - Hadigheh, Ali
N1 - Publisher Copyright:
© 2023 The Author(s)
PY - 2023/12/8
Y1 - 2023/12/8
N2 - When exposed to environmental conditions, fibre-reinforced polymer (FRP) composites are prone to material degradation. The environmental reduction factor in different structural codes reflects the significant effect of the long-term durability of FRPs in aggressive environments. Traditional prediction methods rely on oversimplified premises, which may result in erroneous errors. Due to its proficiency in dealing with complex non-linear structural problems, machine learning (ML) offers a unique potential to increase the predictability of structural engineering factors. This can be attributed to the recent advancements in ML techniques, which leverage their robustness when handling large datasets, as well as the increased processing power that facilitates more efficient data analysis. This article reviews the current implementation cases and capabilities of ML algorithms in overcoming the shortcomings of conventional models for predicting the durability performance of FRP systems. According to the literature, it was found that the efficiency of ML approach varies significantly depending on the quality and comprehensiveness of the database. While various researcher-employed algorithms generally yield accurate predictions for retaining mechanical properties in FRP composites with minor errors, sensitivity analysis highlights varied impacts of variables when using different datasets or machine learning algorithms. This variance may arise from factors like inadequate or low-quality datasets, insufficient training, overfitting, and other influences. More experimental data are needed to enhance the current database to effectively apply ML in more applications for FRPs under different loading and environmental conditions. The paper ends by suggesting future research directions in this field.
AB - When exposed to environmental conditions, fibre-reinforced polymer (FRP) composites are prone to material degradation. The environmental reduction factor in different structural codes reflects the significant effect of the long-term durability of FRPs in aggressive environments. Traditional prediction methods rely on oversimplified premises, which may result in erroneous errors. Due to its proficiency in dealing with complex non-linear structural problems, machine learning (ML) offers a unique potential to increase the predictability of structural engineering factors. This can be attributed to the recent advancements in ML techniques, which leverage their robustness when handling large datasets, as well as the increased processing power that facilitates more efficient data analysis. This article reviews the current implementation cases and capabilities of ML algorithms in overcoming the shortcomings of conventional models for predicting the durability performance of FRP systems. According to the literature, it was found that the efficiency of ML approach varies significantly depending on the quality and comprehensiveness of the database. While various researcher-employed algorithms generally yield accurate predictions for retaining mechanical properties in FRP composites with minor errors, sensitivity analysis highlights varied impacts of variables when using different datasets or machine learning algorithms. This variance may arise from factors like inadequate or low-quality datasets, insufficient training, overfitting, and other influences. More experimental data are needed to enhance the current database to effectively apply ML in more applications for FRPs under different loading and environmental conditions. The paper ends by suggesting future research directions in this field.
KW - Artificial intelligence
KW - FRP durability
KW - Machine learning
KW - Service-life
UR - http://www.scopus.com/inward/record.url?scp=85174357356&partnerID=8YFLogxK
U2 - 10.1016/j.conbuildmat.2023.133692
DO - 10.1016/j.conbuildmat.2023.133692
M3 - Review article
SN - 0950-0618
VL - 408
SP - 1
EP - 16
JO - Construction and Building Materials
JF - Construction and Building Materials
M1 - 133692
ER -