Session: 01-05-01: Applications of Artificial Intelligence/Machine Learning for Aerospace Structures

Paper Number: 107410

107410 - Optimization of a Creased, Thin Elastic Plate to Maximize Flexural Stiffness 

Creased or folded thin plates show higher flexural stiffness compared to that of an equivalent flat sheet of the same material and dimensions. A trivial example for the same can be inferred from a thin sheet of standard A4 size paper that would usually hang under its own weight when supported from the edge of a table; ceases to deform when simply crumpled. This is attributed to the finite curvature at the creases which forces the structure to spend more energy to then deform. In the case of a standard A4 paper, one can observe that simply crumpling the paper did not increase the total mass, but significantly improved the stiffness. On similar lines, a study is conducted to numerically model a thin creased plate of plastic that is 3D printed. As opposed to the paper, this would mean that the printed structure is slightly heavier than an equivalent flat plate. If one were to print such a structure with a uniform creasing, for example, a thin plate with shallow repeated pyramidal units, the creased plate shows up to 146% higher fundamental frequency compared to a flat sheet that is similarly printed. This significant improvement in flexural stiffness compared to a flat sheet comes with less than 1% increase in the total mass. Further, an optimization problem is proposed for a similar structure, where the crease lines are arbitrary. A  flat sheet with two opposite clamped edges and two opposite simply supported edges is considered. The optimization problem is posed as, under constraints of the design, the best way to manufacture a thin elastic sheet with creases, that would maximize the fundamental frequency. A detailed algorithm to create Finite Element (FE) models for recurring iterations suitable for an optimization problem is written using MATLAB and python scripting utility in Abaqus, a commercial FE platform. Further, to investigate the global optimum with reduced computational cost, a surrogate model using Artificial Neural Networks (ANN) is used in conjunction with Genetic Algorithm (GA). The resulting optimal design shows a 176% increase in stiffness compared to the flat sheet (and a 23% increase compared to an ordered creased sheet) with only 0.84% increase in total mass. 

Presenting Author: Avinkrishnan Ambika Vijayachandran University of michigan

Presenting Author Biography: Dr. Avinkrishnan Vijayachandran (Avin Vijay) obtained his Ph.D in Aerospace Engineering from the University of Michigan. He is currently a postdoctoral Research Fellow at the Department of Aerospace Engineering at University of Michigan. Prior to his doctoral studies, he worked as a Future Mobility Researcher at Toyota Research Institute, North America. He obtained his dual Masters in Aerospace Structures and Civil Engineering Structures from the University of Michigan. Dr Vijay's research focuses on Design for Manufacturing, Structural Optimization, Structural stability, Experimental mechanics and machine learning applications in mechanics.

Authors:

Avinkrishnan Ambika Vijayachandran University of michigan
Othman Oudghiri-Idrissi University of Michigan
Andrea Poli University of Michigan
Xiaoming Mao University of Michigan
Ellen Arruda University of Michigan
Serife Tol University of Michigan
Anthony Waas University of Michigan