Title: Integrated approach for an artificial intelligence-based generative product design

Authors: Prasad Umakant Bhad; Rajesh B. Buktar

Addresses: Department of Mechanical Engineering, Sardar Patel College of Engineering, Andheri, Mumbai, 400058, India ' Department of Mechanical Engineering, Sardar Patel College of Engineering, Andheri, Mumbai, 400058, India

Abstract: Generative design is an artificial intelligence-based technology that provides solution with less human intervention. In this research study, we have adopted an integrated approach of convolutional neural network, topology optimisation and autoencoder for generating multiple novel outcomes for one design problem. To showcase the research work, we initially generated bicycle frame designs by providing a reference design to topology optimisation tool to generate optimised designs based on applied constraints, support conditions, loading conditions and environmental conditions. Generated design outcomes were then provided to CNN for image processing. Processed images from CNN were further provided to autoencoder for error less results and dimensional mapping of design domain. Generated outcomes were then converted to 3D models and analysed. Generated designs were validated with yield or critical values of material. The results obtained were within the critical limits of yield values of applied material.

Keywords: artificial intelligence; generative design; topology optimisation; extruded geometric component; convolutional neural network; CNN; deep learning.

DOI: 10.1504/IJDE.2021.124087

International Journal of Design Engineering, 2021 Vol.10 No.2, pp.110 - 135

Received: 22 Jun 2021
Accepted: 22 Jan 2022
Published online: 12 Jul 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article