Session: DFMLC-05-01: Special Session: Design Tool Showcase & Design for Manufacturing and the Life Cycle in response to COVID-19

Paper Number: 68109

Start Time: August 19, 11:10 AM

68109 - An Automated CAD System for the Mass-Customisation of Parametric Models 

Additive Manufacturing enables Mass Customisation of wearables, personal protective equipment, and prosthetics. A wearable product customised to the user could improve comfort, reliability, and performance. However, generating computer aided design (CAD) models for high levels of customisation is often a laborious process, which can result in excessive labour costs and limited scalability. Methods such as parametric modelling reduces the number of operations involved in the CAD model generation process for low to medium degrees of customisation (small number of attributes and features) but cannot adapt dynamically with the increase in the number of geometric features. Professional CAD technicians can create models defined by any number of geometric features but the processing time for each model is long and the accuracy of the models may not always be consistent.

This presentation will propose an automated CAD pipeline for generating fully customised parametric models. The pipeline requires three inputs: a pre-designed parametric CAD model fitted to a generic template of a human body part (i.e., the face); a 3D template (Template Mesh); and a 3D scan of the target body part of a user (Input Mesh). The pipeline output is a CAD model of the product fitted to the user’s body part.

The pipeline is composed of two processes: alignment and fitting; and can be described in full in six steps:

1.      Coarse alignment.

2.      Fine alignment.

3.      Region extraction.

4.      Mesh fitting.

5.      Contact surface generation.

6.      Model generation.

Feature extraction and 3D mesh registration methods such as singular value decomposition (SVD), rigid and nonrigid iterative closest points (ICP, NRICP) are used in the pipeline. More details in strategies for the alignment process, performance optimisation of the fitting step, and potential errors generated at each stage of the pipeline will be discussed in the presentation. In addition, a case study will be presented.

The pipeline was employed to generate respirator masks for a large set of users. Figure 1 shows each stage for the respirator mask custom-fitted to a user’s face. An average runtime of 3 minutes per model, which is significantly less than the equivalent manual process, was recorded. The results of chi-square test of independence tested on 180 user-submitted 3D scans were shown to be independent of the subject’s demographic backgrounds. Overall, the proposed system enables mass-customisation of CAD models and can be adapted to interface with different software applications to construct automated the CAD systems

Presenting Author: Yongxuan Tan Imperial College London

Authors:

Yongxuan Tan Imperial College London
Shiya Li Imperial College London
Mohanad Bahshwan Imperial College London
Joseph Folkes Imperial College London
Samuel Willis Imperial College London
Connor Myant Imperial College London