Design analytics for product family optimization

Han, Hyeongmin

Permalink

https://hdl.handle.net/2142/108151 Copy

Description

Title

Design analytics for product family optimization

Author(s)

Han, Hyeongmin

Issue Date

2020-05-05

Director of Research (if dissertation) or Advisor (if thesis)

• Thurston, Deborah
• Wang, Pingfeng
• Ho, Koki

Doctoral Committee Chair(s)

Kim, Harrison

Department of Study

Industrial&Enterprise Sys Eng

Discipline

Industrial Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Date of Ingest

2020-08-26T23:58:37Z

Keyword(s)

• Design analytics
• Product family
• Mixed-integer programming

Abstract

In competitive markets, manufacturing companies build a large variety of products to satisfy various customer needs. Although the strategy called mass customization enables the company to increase sales in different markets, they may lose the efficiency of mass production. To tackle the challenge, companies have developed product family design. By designing product family, product variants in the family maintain the degree of commonality by sharing elements while producing various types of products. Product family design has received much attention from researchers, but there are rooms for improvements regarding complex systems and integrated design with the supply chain. In this work, I propose a sampling technique that deals with the complexity and non-linearity of performance functions. In engineering design problems, performance functions evaluate the quality of designs. Among the designs, some of them are classified as good designs if responses from the performance functions satisfy design targets. In the early stage of design processes, finding a solution space in the design variable domain or a design exploration is an important procedure to sample well-performing and reliable design candidates. In this thesis, I propose a new method that finds a finite subset of the solution space. The method formulates the problem as optimization problems and utilizes a derivative-free method. With the design sample that is collected from the solution space, a mixed integer linear programs are built and solved iteratively to select the best product family design among the sampled points. The problem is formulated as a weighted set cover problem with a general cost function. As the classical weighted set cover problem which has constant cost can be solved with the greedy algorithm, the modified greedy algorithm for product family design is proposed in this work. The methods are tested to maximize shared components among product family designs in automotive vehicle design. The last work of the thesis includes the supply chain management of the product family design. As the problems in supply chain management are formulated in network models, I propose a network model with module instances and product instances. By combining the proposed structure with the network system of the supply chain, the optimization problem can be obtained. The problem is tested with different scenarios to validate the model.

Graduation Semester

2020-05

Type of Resource

Thesis

Permalink

http://hdl.handle.net/2142/108151

Copyright and License Information

Copyright 2020 Hyeongmin Han

Owning Collections