TY - GEN
T1 - Artificial Neural Network Modeling of FRP-Reinforced Deep Beams
AU - Rashti, Offri
AU - Eid, Rami
AU - Greenberg, Shlomo
AU - Gal, Erez
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.
PY - 2026/1/1
Y1 - 2026/1/1
N2 - The demand for Fiber Reinforced Polymer (FRP) bars in reinforced concrete (RC) members has substantially grown in recent decades due to FRP’s superior durability, high strength-to-weight ratio, and corrosion resistance. This paper suggests using machine learning approach to addresses the design and analysis of a simply supported deep beam with FRP bars. With a comprehensive synthetic database of 200,000 simply supported deep beam configurations that was generated based on ACI & Eurocode 2 strut-and-tie approach to explore the load path method. Several Artificial Neural Network (ANN) algorithms, including feedforward networks and ensemble methods, were trained and evaluated. The findings revealed exceptionally high predictive accuracy, achieving errors consistently below 1.7% for internal forces and below 0.1% for geometric outputs comparing to a new unseen data. From a practical standpoint, the ability of these ANN models to generate highly accurate predictions in near-real time can significantly streamline the design process, particularly when applied to FRP-reinforced deep beams, whose corrosion resistance and strength-to-weight ratio already offer notable advantages. The proposed approach demonstrates significant potential for rapid, automated structural analysis, improving both the accuracy and efficiency of FRP-reinforced deep beam design.
AB - The demand for Fiber Reinforced Polymer (FRP) bars in reinforced concrete (RC) members has substantially grown in recent decades due to FRP’s superior durability, high strength-to-weight ratio, and corrosion resistance. This paper suggests using machine learning approach to addresses the design and analysis of a simply supported deep beam with FRP bars. With a comprehensive synthetic database of 200,000 simply supported deep beam configurations that was generated based on ACI & Eurocode 2 strut-and-tie approach to explore the load path method. Several Artificial Neural Network (ANN) algorithms, including feedforward networks and ensemble methods, were trained and evaluated. The findings revealed exceptionally high predictive accuracy, achieving errors consistently below 1.7% for internal forces and below 0.1% for geometric outputs comparing to a new unseen data. From a practical standpoint, the ability of these ANN models to generate highly accurate predictions in near-real time can significantly streamline the design process, particularly when applied to FRP-reinforced deep beams, whose corrosion resistance and strength-to-weight ratio already offer notable advantages. The proposed approach demonstrates significant potential for rapid, automated structural analysis, improving both the accuracy and efficiency of FRP-reinforced deep beam design.
KW - Artificial Neural Network – ANN
KW - Deep beam
KW - Fiber-reinforced polymer - FRP
KW - Strut-and-tie model
KW - reinforced concrete
UR - https://www.scopus.com/pages/publications/105031492008
U2 - 10.1007/978-3-032-09399-8_65
DO - 10.1007/978-3-032-09399-8_65
M3 - Conference contribution
AN - SCOPUS:105031492008
SN - 9783032093981
T3 - Lecture Notes in Civil Engineering
SP - 691
EP - 702
BT - 12th International Conference on FRP Composites in Civil Engineering, CICE 2025 - Volume 1
A2 - Correia, João R.
A2 - Gonilha, José
A2 - Firmo, João
A2 - Garrido, Mário
A2 - Cabral-Fonseca, Susana
PB - Springer Science and Business Media Deutschland GmbH
T2 - 12th International Conference on Fibre-Reinforced Polymer (FRP) Composites in Civil Engineering, CICE 2025
Y2 - 14 July 2025 through 16 July 2025
ER -