Computational Mechanics for the Twenty-First Century
Edited by: B.H.V. Topping

Chapter 23

NEURAL NETWORKS IN STRUCTURAL ENGINEERING: SOME RECENT RESULTS AND PROSPECTS FOR APLLICATIONS

Z. Waszczyszyn
Institute of Computer Methods in Civil Engineering, Cracow University of Technology, Cracow, Poland

Basics of four types of Artificial Neural Networks (ANNs) related to Error Back-Propagation, Radial Basis Functions, Adaptive Fuzzy Inference System and Recurrent NN (Hopfield-Tank NN) are presented in short. Then applications of these networks in the analysis of the following problems are discussed: 1) analysis of implicit constitutive relations, 2) hybrid FEM/ANN programs, 3) real building vibration problems, 4) damage identification problems. Special attention is paid to accuracy of neural approximation, input data preprocessing and improvements in ANN structure. In the end some conclusions and prospects of neurocomputing in structural engineering are pointed out.

return to the contents page