Abstract
A neural procedure was formulated in [4] as BPNN (Back-Propagation Neural Networks) for the simulation of generalized RMA (Return Mapping Algorithm). This procedure was evaluated to be too large to make a corresponding hybrid FEM/BPNN numerically efficient. That is why two new procedures NPI and NP2 were formulated. A description of their efficiency is presented in the paper, related to the computation number of computer operations and CPU time, carried out by FEM program FEAP and two hybrid programs FEAP/NP1 and FEAP/NP2.
References
[1] T . Furukawa, C. Yagawa. Implicit constitutive modelling for viscoplasticity using neural networks, Int. J. Num. Meth. Eng. , 43: 195- 219, 1998.
[2] Z. Waszczyszyn, E. Pabisek. Hybrid NN/ FEM analysis of elatoplastic plane stress problem, Comp. Assisti. Mech. Eng. Sci ., 6: 177- 188, 1999.
[3] J.C. Simo, T.J .R. Hughes. Computational Inelasticity, Springer-Verlag, New York, 1998.
[4] Z. Waszczyszyn, E. Pabisek. Neural network supported FEM analysis of elastoplastic plate binding, Research News, Budapest UTE, Special Issue 2000/4: 12- 19, 2000.
[5] Z. Waszczyszyn, Cz. Cichoń, M. Radwańska. Stability of Structures by Finite Element Methods, Elsevier, Amsterdam, 1994.
[2] Z. Waszczyszyn, E. Pabisek. Hybrid NN/ FEM analysis of elatoplastic plane stress problem, Comp. Assisti. Mech. Eng. Sci ., 6: 177- 188, 1999.
[3] J.C. Simo, T.J .R. Hughes. Computational Inelasticity, Springer-Verlag, New York, 1998.
[4] Z. Waszczyszyn, E. Pabisek. Neural network supported FEM analysis of elastoplastic plate binding, Research News, Budapest UTE, Special Issue 2000/4: 12- 19, 2000.
[5] Z. Waszczyszyn, Cz. Cichoń, M. Radwańska. Stability of Structures by Finite Element Methods, Elsevier, Amsterdam, 1994.
