Expert Design of Industrial Systems: Formalizing the Design Process Douglas A. Bodner, T. Govindaraj, Karthik N. Karathur, Natalie F. Zerangue, Leon F. McGinnis, Marc Goetschalckx and Gunter P. Sharp Keck Virtual Factory Lab · School of Industrial and Systems Engineering Georgia Institute of Technology Atlanta, GA 30332-0205, USA Abstract A variety of analytic models have been developed by researchers to solve problems in the area of industrial systems design. Examples include storage technology selection models, capacity sizing models, and labor allocation models. Yet practitioners often do not use existing research results, instead relying on expertise and past experience. This research seeks to understand and formalize the process of design as it occurs in practice, and eventually link these results with existing research. Our approach is to study expert designers as they design facilities, focusing on warehouses, with the goal of formalizing design processes and developing methodologies and computational tools to aid in the design process. In this paper, we describe our research methodology, based on the concept of ethnographic studies, and we discuss results obtained to date. These include definitions of functional requirements, specifications of data used, and processes for formulating and evaluating design alternatives. We conclude with a discussion of the types of decision support tools that we are working to create. Keywords Expertise, ethnographic studies, industrial systems design, formalized design process. 1. Introduction Design of industrial systems has been a hallmark of industrial engineering research, focusing on design of factories, warehouses and supply chains (e.g., [12], [13]). This area has witnessed a wealth of published research literature, providing valuable results to solve a number of problems. This paper addresses the relationship between these research results and the practice of design in industry. Here, for the sake of being specific, we focus on warehouse systems. In our interaction with industry practitioners, it has become apparent that there is a disconnect between these research results and the practice of design. Essentially, practitioners, many of whom have accumulated expertise through years of experience, simply do not use most of the extant research. They typically rely on their knowledge from past experiences and on ad-hoc analysis techniques to perform design work. This applies to some extent even to those designers who use a well-established design approach (e.g., Systematic Layout Planning [12]). There are at least two reasons for this. First, from a practical perspective, research results are not widely available as computational tools that can be used by practitioners. Second, and more fundamental, there is a disconnect between the process of design as it is practiced and the formulation of design research results. In the warehousing domain, the literature has concentrated in two main areas: (i) generic frameworks for design (e.g., [1], [17]), and (ii) specific models and tools to solve specific types of problems (e.g., [4], [7], [16], [21]). In general, these two types of research results are not well integrated in the sense that there is a not a direct computational design procedure using both types of results. We might envision such a procedure that would (i) start from one of the generic frameworks, (ii) perform intermediate calculations, (iii) finish by solving specific design sub-problems such as storage technology selection, capacity sizing, layout or labor allocation, and (iv) output a final design. Such a procedure is needed by practitioners. Since it does not exist, the design practitioner faces an uncertain environment when using research results. Hence, designers, who usually operate under significant time constraints, create their own ad-hoc approaches that produce acceptable results. What can be done to address this disconnect? Since research results should be useful to the design practitioner community, the obvious answer is to create a formalized warehouse design methodology that incorporates both types of research results. This is a formidable objective. An alternative approach is to understand and formalize the process of design as it is practiced by industry experts, and then link this formalization with existing and new research to form a design methodology. This paper explores this second approach to the problem. As a starting point, we know that the