Research Area:  Machine Learning

Abstract:

Designing a complex system generally requires its decomposition into smaller modular constituents for the ease of design, integration, operation, and future upgrades. Typically, system decomposition analysis is conducted using an abstract system model during the initial system design stage. The design structure matrix (DSM) has been a popular tool for abstract system modeling. In the DSM, systems are modeled and decomposed into modular configurations using published clustering algorithms. In this paper, we present a novel approach for system architecture decomposition and modularization based on the DSM, which can incorporate multiple design constraints into clustering algorithms. Two clustering algorithms are modified to accommodate the system design constraints. The modified algorithms are used to decompose a static inverter (SIV) of an electrical train into modular configurations that include various design constraints. Finally, the results and analysis of the SIV modularization using new modified clustering algorithms are presented.

Keywords:  
Design Structure Matrix
Integration
Clustering Algorithms
Decomposition
Modularization

Author(s) Name:   Kaushik Sinha, Seok-Youn Han, Eun Suk Suh

Publisher name:  Wiley Online Library

DOI:  10.1002/sys.21518

Volume Information:  Volume 23

Paper Link:   https://incose.onlinelibrary.wiley.com/doi/abs/10.1002/sys.21518