Trade-Off Analysis of System Architecture Modularity Using Design Structure Matrix

Product modularity has been the subject of considerable research and debate in last decade. Various metrics have been proposed in design community to measure the level of modularity and various procedures have been developed to search for ideal modular architectures. These procedures are based on either manual heuristics or computer clustering algorithms. Both approaches are aimed at finding more ideal architectures by optimizing a definition of modularity. However, different desirable criteria are often in conflict with each other and improving one criteria is not feasible without a compromising effect on another. Here, we propose a procedure to find non-dominated optimal architectures where our criteria of interest are intra-cluster and extra-cluster costs. We demonstrate an approach where a designer can consider the architecture that minimizes total cost of interactions, but also allows visualization of the trade-off in increased and decreased costs when considering nearby architectures with more or less modules. An alternative approach has been to consider granularity and hierarchical clustering schemes. We also show through an example that cost optimal architectures for any choice of number of modules are not necessarily obtainable via dividing or aggregating modules, and restricting to hierarchical clustering algorithms produces non-optimal solutions at different numbers of modules.