Applications of Clustering in the Design of Products with Emphasis on Automotive Embedded Systems

Modern design and manufacturing require interdependence for the best quality and cost reduction efforts. A modern design becomes optimal when it is functional, easily achieved, inexpensive, and enables the usage of flexible manufacturing processes. The competitive edge is achieved by delivering predictably good products on a timely basis. The relationship of the components in product design or functions may be easily represented by incidence matrix formulation. Once an incidence matrix is defined, it is useful to rearrange these relationships into clusters to categorize information into sub-units. The clustering in the matrix, although easily done for smaller applications, becomes time consuming for larger applications. By using clustering techniques, design of products can be performed more efficiently.

The computer tool presented in this research paper provides an automated way to perform clustering among interacting elements in a system. An algorithm is introduced that analyzes an input matrix of interactions and then outputs the suggested clustering layouts as an efficient method of maximizing interactions inside the clusters. The tool utilizes the matrix row/column swapping technique similar to Kusiak's Cluster Identification (CI) algorithm [2]. The software is written in MS Visual Basic (VB) and utilizes user-friendly interfaces. In particular, the interactions that can be analyzed are those among components or functions (function-to-function); and, for demonstrative purposes, this paper focuses on embedded system products in the automotive realm.