A neural network approach for predicting the structural behavior of concrete slabs

In this paper the use of a nontraditional technique, neural networks, has b een investigated as a means to develop efficient predictive models of the s tructural behavior of concrete slabs. The applicability of neural networks within the realm of structural analysis is first reviewed, along with their state-of-the-art applications and research efforts. Four neural networks h ave then been developed to model the load-deflection behavior of concrete s labs, the final crack-pattern formation, and both the reinforcing-steel and concrete strain distributions at failure. The four neural networks were tr ained and tested using the experimental results of 38 full-scale slabs. Det ails regarding data modeling, neural network training, and performance eval uation are described. Using the developed networks, a spreadsheet tool for the structural analysis of concrete slabs was developed with a simple inter face, automated predictions, and what-if capabilities. The developed tool i s useful for teaching purposes and for reasonable prediction of the behavio r of concrete slabs without additional experimental testing.