An information theoretic method for developing modular architectures using genetic algorithms

Designing modular products can result in many benefits to both manufacturers and consumers. The development of modular products requires the identification of highly interactive groups of elements and arranging (i.e., clustering) them into modules. However, no rigorous clustering technique can be found in engineering design literature. This paper uses the design structure matrix (DSM) to visualize the product architecture and to develop the basic building blocks required for the identification of product modules. The DSM architectural representation and building blocks are then used for the development of a new clustering method based on the minimum description length (MDL) principle and a simple genetic algorithm (GA). The new method is capable of partitioning the product architecture into a set of modules where interactions within modules are maximized and interactions outside modules are minimized. We demonstrate the proposed clustering method using several examples of real complex products and compare our results to clustering arrangements proposed by human experts. The proposed method is capable of mimicking the clustering preference of human experts and yields competitive clustering arrangements.