Evaluation of the Applicability of Random Walks for Generation of Material Flow Network Models

Modeling material flow systems as networks in terms of graphs is a straightforward approach to investigate these systems. In this context, the nodes represent the workstations and the edges indicate a material flow between two nodes. Furthermore, the individual edges are weighted by the amount of material flow that actually takes place between two workstations. The network model is simple but powerful, because it is easy to build and understand, but it also contains the main information required for the analysis of flow in manufacturing systems. Due to the lack of availability of material flow data from industry, it is sometimes necessary to replicate given real networks in order to perform network based research in manufacturing. The majority of available approaches in the literature generate a network by iteratively adding nodes until the desired network size is reached. In most cases the edge weight is not considered, thus there is no information about the intensity of material flow. However, in the context of manufacturing systems there is a movement of individual parts through the system. These parts have a specific order in which they have to be processed. It is therefore essential to simulate the movement of the parts as a whole. For this purpose, we use the concept of random walks. Random walks describe the movement from a random start node to a random end node, whereby the choice of the subsequent steps is also random. We modify the procedure of random walks so that the properties of these walks are based on the characteristics of real material flow networks, such as number of parts (number of random walks), average number of operations (number of nodes visited by a random walk), etc. In this paper, we show how random walks can be used to generate weighted networks, which exhibit similar properties like real material flow networks. Finally, we evaluate our approach by comparing the properties of real material flow networks and the networks generated using random walks.