A data mining and graph theoretic approach to building generic bills of materials Carol J. Romanowski Rakesh Nagi Department of Industrial Engineering University at Buffalo Buffalo, NY 14260 Abstract Generic bills of materials (GBOMs) are useful as configurators and aid in suggesting new product variants. However, current GBOM research does not address the many problems inherent in building GBOMs from legacy data. This research proposes a data mining and graph theoretic approach that partially automates the process. The resulting GBOMs are represented as constrained XML files that incorporate design and manufacturing information such as alternative parts and cost. Keywords Data mining, generic bills of material, graph theory, engineering design 1. Introduction Mass customization results in many variations on the same product. In some cases, variants are targeted to specific customers, markets, or seasons. For instance, the candy industry often has special wrappers for major holidays such as Christmas or Easter. In other cases, companies provide the semblance of uniqueness by allowing customers to configure their products from an array of available choices. For example, one can specify body color, upholstery color and type, engine type, accessory packages, wheel covers, and other options on a new automobile – thus producing near-one-of-a-kind cars. Computer buyers also can configure “unique” systems by choosing different data storage, display, or multimedia options. The outcome of this customization is a proliferation of BOMs that have relatively minor differences in components or structure, but are in the same product family. Companies need ways to control this proliferation, easily configure product options for customers, and suggest new product variants. In this research, we propose a methodology to address this problem by partially automatic generation of generic bills of material. In our methodology, we make use of graph theory, data mining clustering techniques, and constrained XML. We show how the generic bills are part of a component library that enables engineering designers to quickly find and reuse designs for similar items and parts, thus reducing the effort and time needed in creation of a new product. In the remainder of this paper, Section 2 discusses generic bills of material, graph theory as it relates to trees, clustering algorithms, and constrained XML. Section 3 proposes our methodology for generating GBOMs. In Section 4, we discuss implementation of the methodology and identify areas of future research. 2. Background and literature review 2.1 Generic bills of material (GBOMs) Introduced by Hegge and Wortmann in 1991 [1], generic bills of material (GBOMs) are a method of encapsulating variant design options and alternate components in a single bill. Hegge’s idea of a single specification for all variants of a product family was intended to avoid redundancy while maintaining structural information for assembly, design, and servicing needs. Jiao et al [2] extend Hegge’s GBOM, marrying the generic bill with routing information to form the generic bill-of- materials-and operations (BOMO). Building on Jiao and Tseng’s [3] generic variety structure, the authors propose an object-oriented implementation of the BOMO. Their approach allows for alternate operations but not alternate parts. They assume a common structure among product family variants.