Every now and then cautionary stories emerge: design teams that were not quite as integrated as they should have been - something has been designed partly in centimetres and partly in inches; the redesign that throws up a host of other unforeseen design changes. In these cases, hindsight is a wonderful thing, but how can design teams avoid such pitfalls in the first place?
Product architecture is the result of breaking down the product into systems and systems into components - complex products are made up of systems that are treated as independent modules through much of the design process. Modularity in products and systems is defined in terms of the number of interfaces between modules, and how loosely coupled they are, but it has not yet been explored at the component level.
In this working paper, Manuel E. Sosa, assistant professor of technology and operations management at INSEAD, Steven D. Eppinger of the Sloan School of Management, MIT, and Craig M. Rowles of Pratt & Whitney Aircraft, East Connecticut, bring together a novel combination of product architecture, social networks and graph theory to build a definition of modularity for components. They explore their definitions through the design dependencies and redesign of the Pratt & Whitney PW4098, a large commercial aircraft engine.
The authors define component modularity as the level of independence of one component from others within a product and infer that components become less independent, or more modular, based on design dependencies that describe their connections with other components.
They describe how these connections can be used to measure component modularity and show how social network and graph theory reveal three different types of modularity measures: degree (which relates to the number of directly connected components); distance (which considers design dependency propagation through intermediary components); and bridge (which captures the extent to which a component bridges connections between other components).
Armed with this model, the authors are then able to describe how component-level modularity can be referred to component-level impacts.
This inevitably leads to a number of further questions. Can we assume that various design dependencies are independent of one another? What relative weight should each design dependency receive? Are more modular components less likely to fail than less modular components? Are modular components more or less likely to be redesigned?
Access to the design teams behind the Pratt & Whitney PW4098 gives the authors a wealth of information with which to explore their ideas and examine these questions in detail. They identify five design dependencies at the interfaces between components - spatial, structural, material, energy and information. They examine the product architecture and the design and integration teams behind it, and use their definitions to calculate the modularity measures for the aircraft's engine components.
The results not only help the authors to identify the most modular component, but also enable them to explore the performance-related attributes of the different components and show how managers can use this information to inform their decisions about outsourcing, obsolescence and redesign, especially the allocation of design changes among the various components in a product.
This paper highlights the importance of considering both dependency structure and design change when developing complex products. It opens the door to research in computer-based engineering tools as well as further work into engineering design using combinations of product architecture and social network analysis.
A Network Approach to Define Modularity of Components in Complex Products
Manuel SOSA, Steven Eppinger, Craig Rowles