NEURAL NETWORKS FOR COMPUTING IN FRACTURE-MECHANICS - METHODS AND PROSPECTS OF APPLICATIONS

Authors
THEOCARIS PS PANAGIOTOPOULOS PD
Citation
Ps. Theocaris et Pd. Panagiotopoulos, NEURAL NETWORKS FOR COMPUTING IN FRACTURE-MECHANICS - METHODS AND PROSPECTS OF APPLICATIONS, Computer methods in applied mechanics and engineering, 106(1-2), 1993, pp. 213-228
Citations number
30
Categorie Soggetti
Computer Application, Chemistry & Engineering",Mechanics,"Computer Applications & Cybernetics","Engineering, Mechanical
Journal title
Computer methods in applied mechanics and engineering → ACNP
ISSN journal
00457825
Volume
106
Issue
1-2
Year of publication
1993
Pages
213 - 228
Database
ISI
SICI code
0045-7825(1993)106:1-2<213:NNFCIF>2.0.ZU;2-Z
Abstract
A neural network is proposed and studied for the treatment of fracture mechanics problems. Both the cases of classical cracks and of cracks involving Coulomb's friction or detachment (unilateral contact) interf ace conditions are considered. For the first case, the Hopfield model is appropriately modified, whereas for the second case, a neural model is proposed covering the case of inequalities. For this model, new re sults generalizing the results of Hopfield and Tank are obtained. Nume rical applications illustrate the theory. Finally, the parameter ident ification problem for fractured bodies is formulated as a supervised l earning problem.