Research on matrix based complexity management has come a long way. Originating from a process focus with the first published formulation of a Design Structure Matrix (DSM) by Don Steward in 1981 , a whole community has developed around this research. The DSM is able to model and analyze dependencies of one single type within one single domain. For a product, e.g. the domain “components” can be regarded. Using the relationship type “change of component 1 causes change of component 2”, an assembly can be analyzed with regard to the overall change impacts in order to model possible change propagations. DSMs can have different qualities: Binary DSMs represent only the existence of a relation, whereas numerical DSMs represent a numerical value (also called “weight”) to represent the strength of a relation. DSMs can either be directed (as shown in the figure below), or non-directed. DSMs are never reflexive, i.e. a relation from an element to the same element is not permissible.
The figure shows a simple process consisting of six tasks that are shown as a flow chart on the right hand side and a DSM representing that process on the left hand side. There are numerous algorithms to analyze the overall structure of the relationships within a DSM, e.g. tearing, banding and partitioning, or the analysis for different structural properties (see DSM tutorial).
The Design Structure Matrix (DSM) 