Applications of Neural Networks in Modeling and Design of Structural Systems

Abstract

There has been considerable recent activity in exploring biological motivated computational paradigms in problems of engineering analysis and design. Such computational models are placed in a broad category of soft-computing tools that span the gap between traditional procedural methods of computation on one side, and heuristics driven inference engines (non procedural methods) on the other. Of these, methods of neural computing and evolutionary search have been extensively explored in problems of structural analysis and design. The purpose of the present chapter is two-fold. It provides an overview of those neural network architectures that are pertinent to the problem of structural analysis and design, including the back-propagation network, the counterpropagation network, the ART network, and the Hopfield network. It then provides a summary of select applications of neurocomputing in the field of structural synthesis. This summary includes the applications of neural networks in function modeling, in establishing causality in design data, in function optimization, and in diagnostics of structural systems.