Title:
Neural Network Modeling of Concrete Expansion during Long-Term Sulfate Exposure
Author(s):
Rami M. Haj-Ali, Kimberly E. Kurtis, and Akshay R. Sthapit
Publication:
Materials Journal
Volume:
98
Issue:
1
Appears on pages(s):
36-43
Keywords:
concrete; durability; expansion; sulfate attack
DOI:
10.14359/10158
Date:
1/1/2001
Abstract:
The development of artificial neural network (ANN) models to predict the long-term expansion response of concrete cylinders while exposed to a 2.1% Na2SO4 sulfate solution is described in this paper. The experimental data used in this study was collected by the U.S. Bureau of Reclamation (USBR) during a long-term (40+ years), nonaccelerated test program. The ANN is constructed such that it provides for the expansion of the concrete as an output while its input vector includes time and two mixture parameters: the water-cement ratio (w/c), and the tricalcium aluminate content of the cement. Different approaches for developing and training the ANN are discussed in this study. The effectiveness of the trained ANN is evaluated by comparing its response with the experimental data that were used in the training. The trained ANN model is also compared to two additional experimental cases that were not used in the pretraining process. It is shown that ANN can effectively learn and predict the expansion of the concrete samples within a practical range of the two mixture parameters, during a span of up to 40 years, despite the described limitations of using the USBR data for ANN training.