An anticipated shear design method for reinforced concrete beams strengthened with anchoraged carbon fiber-reinforced polymer by using neural network

Tanarslan H. M., Kumanlioglu A., Şakar G.

STRUCTURAL DESIGN OF TALL AND SPECIAL BUILDINGS, vol.24, no.1, pp.19-39, 2015 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 24 Issue: 1
  • Publication Date: 2015
  • Doi Number: 10.1002/tal.1152
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.19-39
  • Keywords: anchorage, shear strengthening, neural networks, FRP, RC BEAMS, CAPACITY, PERFORMANCE
  • Dokuz Eylül University Affiliated: Yes


Using externally bonded carbon fiber-reinforced polymer (FRP) for strengthening has been turned into a popular decision owing to its mechanical leads. Consequently, design guidelines and researchers have established several analytical equations to predict the contribution of FRP to ultimate shear capacity. The developed analytical equations projected the influence of FRP reinforcements within certain limits. However, not mentioned parameters such as the shear span-to-depth ratio and anchorage application influence the ultimate behavior of strengthened specimens. Accordingly, distant predictions between test results and code predictions are observed for the specimens in whom anchorage is applied. As an alternative method, artificial neural network (NN) can be used to predict the contribution of anchoraged carbon FRP to shear strength of deficient reinforced concrete beams. Accordingly, two NN models with back-propagation are developed in this study. Unlike the existing design codes, the model considers the effect of anchorage and the shear span-to-depth ratio at the ultimate state. Artificial NN model is trained, validated and tested using the literature of 79 reinforced concrete beams. Then, NN results are compared with those theoretical' predictions calculated directly from International Federation for Structural Concrete, the American guideline (ACI 440.2R) and the Australian guideline. Within all theoretical predictions of design guidelines, fib14 provided the best predictions according to experimental results. Consequently, 25% of fib14 predictions are within +/- 10% of the experimental results, and also, 65% of the fib14 predictions are within +/- 25% of the measured values. Besides, executed comparisons indicated that the NN model is more exact than the guideline equations with respect to the experimental results and can be applied effectively within the range of parameters covered in this study. Copyright (c) 2014 John Wiley & Sons, Ltd.