Keywords: bond, reinforced concrete, prestressed concrete, computational intelligence, neural networks, genetic algorithms.

This study aims to produce equations that characterize the bond behavior between reinforcing bars and concrete and between seven-wire strands and prestressed concrete, using genetic algorithms. The available empirical equations are dimensionally inhomogeneous relationships, derived from experimental results on specimens reinforced with prototype bar sizes only. Also, because of the differences in bond strength and the response of small-size bars and wires and large-size bars and seven-wire strands, it is difficult to ensure true modeling of bond. The interest in this work, rests on the optimization of the available equations in order to reduce the errors between the results of these equations and experimental data.

The bond stress-slip relationship was evaluated from experimental data obtained from pull-out tests [1]. In preceding studies [2] neural networks were trained to evaluate the influence of the geometrical and material parameters of the concrete on the bond stress value. The performance of neural networks was satisfactory. It was verified that a backpropagation neural network with one hidden layer, even with a restricted amount of experimental data [3,4,5], learned the difference between the bond behavior of reinforced and prestressed concrete as well as the variation of the bond stress in function of the bar diameter and the concrete compressive strength. This neural network can generate new data allowing a parametric analysis of the problem.

Neural networks do not supply mathematical explanations of the bond stress, due to the training algorithms characteristics and architecture complexity. However, their results can be used as the objective to the optimization of formulas. In this work, genetic algorithms were adopted for the creation of a mathematical formula that represents the results found by the neural network.

Starting from bibliographical models restricted to slip limits and test specifications, it was possible to model equations that can represent concrete with several bar diameters that is also valid for sliding between zero and the maximum sliding of the respective concrete types. This work will continue in order to find other equations to characterize high-performance concrete and prestressed concrete.

1

Standard RILEM/CEB/FIP-Recommendation RC-6, "Bond test for Reinforcement Steel-Pull-Out Test", Concrete Reinforcement Technology, Paris, 1983.

2

M.E.N. Tavares, E. Tominaga, E.T. Fonseca, "Avaliação da resistência à aderência das natas de cimento utilizando redes neurais", Simpósio Ibero-Americano - Siabe, 2005.

3

M.E.N. Tavares, "Ensaios de arrancamento em barras nervuradas", Seminário de Doutorado, Coppe/Ufrj, Brasil, March 1997.

4

M.T.G. Barbosa, "Avaliação do Comportamento da Aderência em Concretos de Diferentes Classes de Resistência", Tese de Doutorado, Coppe/Ufrj, Brasil, October 2001.

5

CEB, "Bond of Reinforced Bars with Small Diameters Determined by Pull-Out Test", In: Bulletim D'Informations 186, Paris, 59-96, June 1988.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £62 +P&P)