Multi-objective shape optimization of plate structure under stress criteria based on sub-structured mixed FEM and genetic algorithms

11th International Conference on Damage Assessment of Structures

物理学;材料科学

Garambois, Pierre^1 ; Besset, Sebastien^1 ; Jézéquel, Louis^1

Laboratoire de Tribologie et Dynamique des Systèmes, UMR CNRS 5513, Ecole Centrale de Lyon, 36 avenue Guy de Collongue, Ecully cedex

69134, France^1

关键词: Different thickness;    Linear dependence;    Mixed finite elements;    Pareto optimal solutions;    Reduction method;    Structure analysis;    Thickness distributions;    Thickness parameters;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/628/1/012011/pdf
DOI  :  10.1088/1742-6596/628/1/012011

学科分类：材料科学（综合）

来源: IOP

【 摘 要 】

This paper presents a methodology for the multi-objective (MO) shape optimization of plate structure under stress criteria, based on a mixed Finite Element Model (FEM) enhanced with a sub-structuring method. The optimization is performed with a classical Genetic Algorithm (GA) method based on Pareto-optimal solutions and considers thickness distributions parameters and antagonist objectives among them stress criteria. We implement a displacement-stress Dynamic Mixed FEM (DM-FEM) for plate structure vibrations analysis. Such a model gives a privileged access to the stress within the plate structure compared to primal classical FEM, and features a linear dependence to the thickness parameters. A sub-structuring reduction method is also computed in order to reduce the size of the mixed FEM and split the given structure into smaller ones with their own thickness parameters. Those methods combined enable a fast and stress-wise efficient structure analysis, and improve the performance of the repetitive GA. A few cases of minimizing the mass and the maximum Von Mises stress within a plate structure under a dynamic load put forward the relevance of our method with promising results. It is able to satisfy multiple damage criteria with different thickness distributions, and use a smaller FEM.