Design Structure Matrix

from Wikipedia, the free encyclopedia

The Design Structure Matrix (DSM) is a method for recording, modeling, analyzing and synthesizing (within certain limits) the networking of elements in highly networked systems. Typical examples of such systems are complex, highly integrated product architectures, organizational structures or processes. In such systems, the Design Structure Matrix makes it possible to relate elements of one type if they are connected to one another via a comparable type of relationship.

A square matrix is ​​used as the modeling basis, which depicts the individual elements of the system on the vertical and transverse axes, and each individual cell can be used to depict the relationship between two elements. Such a DSM can be modeled as "row influences column" or as "column influences row"; While the first type of modeling has more established itself in the European-speaking area, the latter is particularly used in North America and Asia. The European reading direction is used in the following illustration.

Design Structure Matrix of a process, which is shown as a graph on the right in the picture; the DSM is read as "process step in line starts process step in column"

A Design Structure Matrix can represent the following types of relationships:

  • Binary DSM: Only the existence of a relationship is represented, typically by a 1 or an X in a cell of the matrix.
  • Numerical DSM: The relationship can be weighted; mostly values ​​between 0 and 100% are used, but other weightings are also possible.

No reflexive relationships are mapped here, i. H. an element cannot relate to itself; the diagonal of the matrix is ​​therefore always empty.

Binary DSM

Binary design structure matrix of a process in which the weighting of the individual elements is given in percent (e.g. as an attribute of the duration of a process step or its difficulty in execution)

Numerical DSM

Numerical Design Structure Matrix of a Process; the varying degrees of dependency of one process step on the next is given as a value between 0 and 1

The DSM is traditionally used to set up stringent processes. For this purpose, the methods of partitioning, banding and triangularization have become established. The aim is to identify independently executable parts of the process (so-called "bands") and to transfer the model into a so-called upper triangular matrix, which is equivalent to an ideal sequence with as few returns as possible in the process. In addition, the method of clustering is used to group closely networked groups of process steps as a cluster and thus to ensure rapid processing and to assign adequate teams.

DSM is used in a similar way for product architectures. In such cases, clustering can be used to identify modular architectures (keyword: platform or construction kit). It is also possible to identify the effects of changes to a component by tracking all outgoing paths of the dependencies in the product architecture and thus far-reaching effects can be identified at an early stage.

In a similar way, organizational structures and team structures can also be modeled. The DSM also offers the possibility of handling non-hierarchical structures that are networked "criss-cross".

Further modeling

The restriction that a DSM can only have one type of element, e.g. B. can only map components of a product, can be bypassed with the Domain Mapping Matrix (DMM); this relates two so-called domains to one another. It is also possible to carry out analyzes such as clustering for such matrices.

Example of a domain mapping matrix:

Binary Domain Mapping Matrix

A comprehensive modeling system that can capture any system in its complexity is the multiple-domain matrix . This combines different DSMs and DMMs into a comprehensive model that also allows analyzes across multiple domains and different types of relationships. For this purpose, aggregated DSMs can be determined and used as specific views of a system. Their application to different systems allows a much more targeted handling of complex systems.

Example for a multiple-domain matrix:

Binary multiple-domain matrix that combines the above examples into a complete system

literature

  1. ↑ Stefan Friedrich: Possible uses of a design structure matrix in the context of strategic project management. GRIN Verlag, 2008, accessed on February 18, 2017 .
  2. ^ DV Steward: The Design Structure System: A Method for Managing the Design of Complex Systems. In: IEEE Transactions on Engineering Management. 28 (3), 1981, pp. 71-74.
  3. ^ T. Browning: Applying the Design Structure Matrix to System Decomposition and Integration Problems: A Review and New Directions. In: IEEE Transactions on Engineering Management. 48 (3), 2001, pp. 292-306.
  4. ^ R. Keller: Predicting change propagation: algorithms, representations, software tools. PhD thesis, Cambridge University Engineering Department, 2007.
  5. ↑ M. Danilovic; TR Browning: Managing Complex Product Development Projects with Design Structure Matrices and Domain Mapping Matrices. In: International Journal of Project Management. 25 (3), 2007, pp. 300-314.
  6. ^ M. Maurer: Structural awareness in complex product design. Munich: Dr-Hut 2007.
  7. ↑ U. Lindemann; M. Maurer; T. Braun: Structural Complexity Management. Berlin: Springer 2009.

Web links