Reliability based optimization of laminated composite structures using genetic algorithms and Artificial Neural Networks

► Reliability based design optimization of composite structures with surrogate models. ► Simple to complex examples used to validate the methodology. ► RBDO on 3D composite shell including stochastic fields for layer thickness. ► Acceptable accuracy and reduced processing time for this methodology....

Full description

Saved in:
Bibliographic Details
Published in:Structural safety 2011-05, Vol.33 (3), p.186-195
Main Authors: Gomes, Herbert Martins, Awruch, Armando Miguel, Lopes, Paulo André Menezes
Format: Article
Language:English
Subjects:
Accuracy
Applied sciences
Artificial Neural Networks
Buildings. Public works
Composite materials
Composite structures
Computation methods. Tables. Charts
Exact sciences and technology
Finite element analysis
Genetic algorithms
Mathematical models
Methodology
Monte Carlo methods
Optimization
Reliability analysis
Structural analysis. Stresses
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:► Reliability based design optimization of composite structures with surrogate models. ► Simple to complex examples used to validate the methodology. ► RBDO on 3D composite shell including stochastic fields for layer thickness. ► Acceptable accuracy and reduced processing time for this methodology. ► Improved reliability of composite design by ply angle re-orientation. The design of anisotropic laminated composite structures is very susceptible to changes in loading, angle of fiber orientation and ply thickness. Thus, optimization of such structures, using a reliability index as a constraint, is an important problem to be dealt. This paper addresses the problem of structural optimization of laminated composite materials with reliability constraint using a genetic algorithm and two types of neural networks. The reliability analysis is performed using one of the following methods: FORM, modified FORM (FORM with multiple checkpoints), the Standard or Direct Monte Carlo and Monte Carlo with Importance Sampling. The optimization process is performed using a genetic algorithm. To overcome high computational cost it is used Multilayer Perceptron or Radial Basis Artificial Neural Networks. It is shown, presenting two examples, that this methodology can be used without loss of accuracy and large computational time savings, even when dealing with non-linear behavior.
ISSN:0167-4730
1879-3355
DOI:10.1016/j.strusafe.2011.03.001