Keywords: artificial neural network, dissipated energy, low-cycle, fatigue life.

The hysteresis energy dissipated in steel reinforcing bars under low-cycle fatigue load is an important parameter in predicting their fatigue life. The fatigue life of steel reinforcing bars and therefore the amount of hysteresis energy dissipated, directly influences the level of damage in reinforced concrete buildings under seismic load. In this study artificial neural networks (ANNs) are used to predict the hysteresis energy dissipated in steel reinforcing bars. Although an ANN has been used before to predict the fatigue life of steel reinforcing bars and other metals [1,2,3], to the best of the authors' knowledge no study has been conducted to predict the hysteresis energy dissipated in steel reinforcing bars under cyclic loading using anANN. Therefore, an ANN is used to predict the hysteresis energy dissipated in steel reinforcing bars subject to low-cycle fatigue. The input parameters for the network include the total strain amplitude and the strain ratio. The output of the ANN is the energy dissipated in the first cycle, the energy dissipated in the average cycles and the total energy dissipated in all cycles to failure. The data used to train and test the ANN are the results of experiments conducted by the authors for different types of steel reinforcing bars, subjected to variable strain amplitudes beyond yield.

The ANN was trained using the experimental data and then tested using the test data to predict the desired hysteresis energy. The ANN predicted and experimentally measured or desired energy values were compared and the accuracy of the ANN was evaluated. It is concluded that:

• The ANN can accurately predict the desired hysteresis energy dissipated in the first cycles, average cycles and total cycles. This in spite of the relatively small amount of data used for training.
• The prediction of the first-cycle energy and the average cycle energy is more accurate than the prediction of the total cycles energy. Prediction of the first-cycle energy is slightly more accurate than the prediction of the average cycles energy.
• Predictions of the energy for most test samples are within 10% of the measured first-cycle energy values, within 15% of the average cycles energy values and within 30% of the total cycles energy values.
• Based on the trained ANN, a parametric study can be carried out as an extension to this work to investigate the effect of a range of strain amplitudes and strain ratios on the amount of energy dissipated at different levels of low-cycle fatigue loading.

1

J.A. Abdalla, R. Hawileh, "Predictions of low-cycle fatigue life of steel reinforcing bars using artificial neural network", Proceedings of the 3rd International Conference on Modeling Simulation and Applied Optimization (ICMSAO'9), Sharjah, UAE, January 20-22, 2009.

2

T.T. Pleune, O.K. Chopra, "Using artificial neural networks to predict the fatigue life of carbon and low-alloy steels", Nuclear Engineering and Design, 197, 1-12, 2000. doi:10.1016/S0029-5493(99)00252-6

3

K. Genel, "Application of artificial neural network for predicting strain-life fatigue properties of steels on the basis of tensile tests", International Journal of Fatigue, 26, 1027-1035, 2004. doi:10.1016/j.ijfatigue.2004.03.009

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 £78 +P&P)