The graft of ANN-FEM technique in Macro-mechanics of Multi-oriented Natural Fiber/Polyester Laminates

Document Type : Research Paper

Authors

1 Department of Mechanical Engineering, Nnamdi Azikiwe University, Awka, Nigeria

2 Department of Industrial and Production Engineering, Nnamdi Azikiwe University, Awka, Nigeria

10.22075/macs.2021.20579.1263

Abstract

Low weight and high strength requirements are prime target design objectives in strength demanding applications. Skillful design of low density, low weight and eco-friendly natural fiber composites could provide an alternative material route to the actualization of lighter structures. The present study proposed ANN-FEM computational framework for the macro-mechanical analysis of multi-oriented Plantain Empty Fruit Bunch Fiber Laminate (PEFBFL) and Plantain Pseudo Stem Fiber Laminate (PPSFL). Control factors were numerically varied using Finite Element Method (FEM) and the resultant FEM models which encapsulated material properties of the laminate was streamlined into Artificial Neural Network (ANN) training scheme. A standard feed-forward backpropagation network was adopted and the ANN model consists of stacking sequence, laminate aspect ratio and fiber orientation as input variables while the selected network outputs variables include average stress and displacement. The laminate constitutive equation was developed which enabled the establishment of laminate load deformation affiliation and equivalent elastic constants. The damage onset for individual lamina was detected by the maximum principal stress theory and the overall laminate strength of 40.12 N/mm^2 was obtained for PEFBFL and 32.16N/mm^2 for PPSFL. On the whole, there was steady reduction in laminates elastic modulus which points to compromised stiffness in material principal axis arising from gradual failure of the plies, this trend continued until the last ply failure occurred in ply 3 and 4 at 90 degrees in tensile mode of transverse direction. Stresses and displacements observed using CLT agree very closely with predictions of ANN.

Keywords