Computational Technology Resources - CCP

Keywords: identification, composite, genetic algorithms, nonlinear model, parallel selection, artificial neural networks.

Summary

Identification of a constitutive model that includes damage, elastoplasticity, viscoelasticity and viscoplasticity will be presented in this article. The description of the model was previously validated in the following references [1,2] in the framework of a meso-macro modelling. In other to perform this task of identification we first have implemented the behaviour of the model in a triangular multi-layer finite element shell in the code CAST3M©. Once the model is implemented, the second task which is the most important consists in developing a robust non linear optimization tool.

Because the identification problem can be formulated as: a minimization or a maximization of one cost function subject to some limitations. The common search algorithms based on steepest gradients fail in general in finding the optimal solution. So, we introduced the idea of the genetic algorithms [3,4] (GAs). The principle of the GAs is to simulate the evolution of one population of individuals to which we apply different production operators. As the GAs start searching from different initial solutions, it gave them a global view of the problem. This global perspective prevents them to be trapped locally and allow them to explore all the research landscape. The behaviour of this algorithm is similar to a black box with entries and one exit.

Among the major advantage of the GAs is the fact that they do not need any starting admissible solution. To decrease the time cost, we have developed a hybrid strategy that use one GA based on parallel selection GAPS coupled with a local search method FFSQP [5], and an approximation method based on artificial neural networks ANNs [6].

The parallel selection method implemented in our program GAPS consists in introducing in a selection one or several individuals coming from others selection processes. This allows the different process to exchange information regarding to the area of research for a better exploration in all the landscape.

As the value of the cost function can generally be estimated only by time consuming simulation, which is very important, we use data, produced from early generation of the GAPS to train the ANN, and then once the ANN is well trained, it's used as an approximation (Meta) models of the real problem for the rest of the identification process.

Finally, the efficiency of the developed methods is applied on a multi-layered polymer based composite structures. The laminate is made of unidirectional layers of polymer matrix reinforced with long fibres.

References

1: M.L. Boubakar, F. Trivaudey, D. Perreux and L. Vang, "A meso-macro finite element modelling of laminate structures, Part I: Time-independent behaviour", Composite Structures, 58: 271-286, 2002. doi:10.1016/S0263-8223(02)00049-1
2: M.L. Boubakar, F. Trivaudey, D. Perreux and L. Vang, "A meso-macro finite element modelling of laminate structures, Part II: Time-dependent behaviour", Composite Structures, 60: 275-305, 2002. doi:10.1016/S0263-8223(03)00012-6
3: J.H. Holland, "Adaptation in natural and artificial systems". Ann Arbor: University of Michigan Press, 1975.
4: D.E. Goldberg, "Genetic algorithms in search, optimisation, and machine learning". New York: Addison-Wesley, 1989.
5: J.L. Zhou and A.L. Tits, User's guide for FFSQP version 3.7: "A Fortran code for solving optimization programs, possibly Minimax, with general inequality constraints and linear equality constraints, generating feasible iterates", Institute for Systems Research, University of Maryland, Technical Report SRC-TR-92-107r5, College Park, MD 20742, 1997.
6: S. Guessasma, G. Montavon, C. Coddet, "Modelling of the APS plasma spray process using artificial neural networks: basis, requirements and an example", Comput. Mat. Sci., 29: 315-333, 2004. doi:10.1016/j.commatsci.2003.10.007

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £80 +P&P)

	Computational & Technology Resources an online resource for computational, engineering & technology publications
	not logged in - login
Front Page Browse CCP CSETS CTR IJRT Other Authors Search Purchase Guide FAQ Contact us	Civil-Comp Proceedings ISSN 1759-3433 CCP: 82 PROCEEDINGS OF THE EIGHTH INTERNATIONAL CONFERENCE ON THE APPLICATION OF ARTIFICIAL INTELLIGENCE TO CIVIL, STRUCTURAL AND ENVIRONMENTAL ENGINEERING Edited by: B.H.V. Topping Paper 37 A Hybrid Computational Strategy for Identification of a Non Linear Composite Model D.H. Bassir+ and S. Guessasma* +Laboratory of Applied Mechanic Raymond Charléat, Institute of FEMTO-ST, University of Franche-Comté / CNRS, Besançon, France +Mechanical Department, Technical University of Belfort-Montbeliard, Belfort, France doi:10.4203/ccp.82.37 purchase the full-text of this paper Full Bibliographic Reference for this paper D.H. Bassir, S. Guessasma, "A Hybrid Computational Strategy for Identification of a Non Linear Composite Model", in B.H.V. Topping, (Editor), "Proceedings of the Eighth International Conference on the Application of Artificial Intelligence to Civil, Structural and Environmental Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 37, 2005. doi:10.4203/ccp.82.37 Keywords: identification, composite, genetic algorithms, nonlinear model, parallel selection, artificial neural networks. Summary Identification of a constitutive model that includes damage, elastoplasticity, viscoelasticity and viscoplasticity will be presented in this article. The description of the model was previously validated in the following references [1,2] in the framework of a meso-macro modelling. In other to perform this task of identification we first have implemented the behaviour of the model in a triangular multi-layer finite element shell in the code CAST3M©. Once the model is implemented, the second task which is the most important consists in developing a robust non linear optimization tool. Because the identification problem can be formulated as: a minimization or a maximization of one cost function subject to some limitations. The common search algorithms based on steepest gradients fail in general in finding the optimal solution. So, we introduced the idea of the genetic algorithms [3,4] (GAs). The principle of the GAs is to simulate the evolution of one population of individuals to which we apply different production operators. As the GAs start searching from different initial solutions, it gave them a global view of the problem. This global perspective prevents them to be trapped locally and allow them to explore all the research landscape. The behaviour of this algorithm is similar to a black box with entries and one exit. Among the major advantage of the GAs is the fact that they do not need any starting admissible solution. To decrease the time cost, we have developed a hybrid strategy that use one GA based on parallel selection GAPS coupled with a local search method FFSQP [5], and an approximation method based on artificial neural networks ANNs [6]. The parallel selection method implemented in our program GAPS consists in introducing in a selection one or several individuals coming from others selection processes. This allows the different process to exchange information regarding to the area of research for a better exploration in all the landscape. As the value of the cost function can generally be estimated only by time consuming simulation, which is very important, we use data, produced from early generation of the GAPS to train the ANN, and then once the ANN is well trained, it's used as an approximation (Meta) models of the real problem for the rest of the identification process. Finally, the efficiency of the developed methods is applied on a multi-layered polymer based composite structures. The laminate is made of unidirectional layers of polymer matrix reinforced with long fibres. References 1 M.L. Boubakar, F. Trivaudey, D. Perreux and L. Vang, "A meso-macro finite element modelling of laminate structures, Part I: Time-independent behaviour", Composite Structures, 58: 271-286, 2002. doi:10.1016/S0263-8223(02)00049-1 2 M.L. Boubakar, F. Trivaudey, D. Perreux and L. Vang, "A meso-macro finite element modelling of laminate structures, Part II: Time-dependent behaviour", Composite Structures, 60: 275-305, 2002. doi:10.1016/S0263-8223(03)00012-6 3 J.H. Holland, "Adaptation in natural and artificial systems". Ann Arbor: University of Michigan Press, 1975. 4 D.E. Goldberg, "Genetic algorithms in search, optimisation, and machine learning". New York: Addison-Wesley, 1989. 5 J.L. Zhou and A.L. Tits, User's guide for FFSQP version 3.7: "A Fortran code for solving optimization programs, possibly Minimax, with general inequality constraints and linear equality constraints, generating feasible iterates", Institute for Systems Research, University of Maryland, Technical Report SRC-TR-92-107r5, College Park, MD 20742, 1997. 6 S. Guessasma, G. Montavon, C. Coddet, "Modelling of the APS plasma spray process using artificial neural networks: basis, requirements and an example", Comput. Mat. Sci., 29: 315-333, 2004. doi:10.1016/j.commatsci.2003.10.007 purchase the full-text of this paper (price £20) go to the previous paper go to the next paper return to the table of contents return to the book description purchase this book (price £80 +P&P)
Back to top	©Civil-Comp Limited 2023 - terms & conditions