A Spectral Analysis Software to Detect Modules in a DSM

DSM 14 Proceedings of the 16th International DSM conference: Risk and Change management in complex systems

Year: 2014
Editor: Marle, F.; Jankovic, M.; Maurer, M.; Schmidt, D. M.; Lindemann, U.
Author: Sarkar, S.; Dong, A.
Series: DSM
Section: System Architecture and Product Modularity
Page(s): 55-64

Abstract

This paper presents a new spectral clustering and partitioning based software tool for the identification of modules in design structure matrices. The MATLAB-based software can identify the globally optimal number of modules in the design structure matrix and identify overlapping and hierarchically overlapping modules. The software tool provides the capability for the modeler to vary the level of granularity of the analysis so as to obtain either a high-level or a granular view of modularity at the components, sub-systems, or system levels, the number of levels defined arbitrarily. It also provides the modeler with the flexibility of visualizing the membership of overlapping components to modules in terms of continuously varying membership strength. A link is provided for interested readers to download the software.

Keywords: Product architecture, modularity, eigenvalue decomposition, spectral clustering