Fibre architecture design of 3D woven composite with genetic algorithms: a unit cell based optimisation framework and performance assessment

Zeng, Xuesen; Long, Andew C.A.; Ashcroft, Ian; Potluri, Prasad

Abstract

There are vast possibilities in fibre architecture design of 3D woven reinforcement. This paper considers the application of Genetic Algorithm (GA) in 3D woven composites optimisation. A set of real and integral variables, representing 3D fibre architecture, are formulated into a mixed integer Genetic Algorithm. The objective function is evaluated through automation of the unit cell based finite element analysis, by using the open source pre-processor TexGen and the commercial solver ABAQUS. The mixed integer Genetic Algorithm is adapted to a micro-population, aiming to improve computational efficiency. The study uses statistical tests to quantify the performance of the Genetic Algorithm schemes and the choice of parameters. The proposed approach was applied to the optimisation of 3D woven composites for maximum buckling resistance for the case of a landing gear brace. This study demonstrated that the optimisation converged to the optimum design within 20 iterations, considering 300 out of 7000 permissible solutions. In terms of buckling performance, the optimum design performed twice as well as cross-ply laminated composites and at least 50% better than known orthogonal 3D woven composites.

Citation

Zeng, X., Long, A. C., Ashcroft, I., & Potluri, P. Fibre architecture design of 3D woven composite with genetic algorithms: a unit cell based optimisation framework and performance assessment. Presented at 20th International Conference on Composite Materials

• Conference Name: 20th International Conference on Composite Materials
• End Date: Jul 24, 2015
• Publication Date: Jul 20, 2015
• Deposit Date: Aug 5, 2015
• Publicly Available Date: Aug 5, 2015
• Peer Reviewed: Not Peer Reviewed
• Keywords: 3D woven composites, design optimisation, Genetic Algorithm, unit cell
• Public URL: https://nottingham-repository.worktribe.com/output/756592
• Publisher URL: http://www.iccm20.org/fullpapers/file?f=zRDv7On0XQ
• Related Public URLs: http://www.iccm20.org/