Complex Networks in Manufacturing - Suitability and Interpretation

The fourth industrial revolution, and the associated digitization of the manufacturing industry, has resulted in increased data generation. Industry leaders aim to leverage this data to enhance productivity, boost innovation and generate new manners of competition. In this work, out of the many domains within the manufacturing sector, production will be explored. To this end, the mathematical tools of network science are utilized to characterize and evaluate production networks in terms of complex networks.

In a manufacturing complex network, nodes represent workstations, and directed edges abstract the material flow that occurs among pairs of workstations. These types of complex networks are known as "material flow networks" and are used to study issues associated with manufacturing systems in the domain of production at the intra-enterprise level. While some research on the subject exists, this work will demonstrate that the use of complex networks to describe and evaluate manufacturing systems constitutes a nascent research field. In fact, the limited existing literature tackles a vast number of issues raising more questions than providing answers.

This work aims to answer a number of those open questions. Firstly, which complex network metrics are suitable in the context of manufacturing networks will be determined. As a consequence, unsuitable metrics will be identified as well. To accomplish this, the flow underlying assumptions of popular complex network metrics is studied and compared to those of manufacturing networks. Furthermore, other existing complex network metrics with more appropriate underlying assumptions, but not yet explored in the context of manufacturing, are proposed and evaluated. Then, the appropriate interpretation of suitable complex network metrics in terms of Operations Research is provided. Finally, shortcomings of these metrics are highlighted to caution practitioners regarding their use in industrial settings.