Complex networks of material flow in manufacturing and logistics: Modeling, analysis, and prediction using stochastic block models. (July 2020)

• Record Type: Journal Article
• Authors: Funke, Thorben; Becker, Till
• Is Part Of: Journal of manufacturing systems. Volume 56(2020)
• Page Start: 296
• Page End: 311
• Publication Date: 2020-07
• Subjects: Manufacturing systems -- Logistics -- Complex networks -- Stochastic block model -- Material flow -- Prediction; Manufacturing processes -- Periodicals; Production engineering -- Data processing -- Periodicals; Robots, Industrial -- Periodicals; Production, Technique de la -- Informatique -- Périodiques; Robots industriels -- Périodiques; Electronic journals; 670.42
• DOI: 10.1016/j.jmsy.2020.06.015
• Languages: English
• ISSNs: 0278-6125

Abstract

Highlights: A concrete and applicable method for the prediction of future states of a material flow network is presented. The underlying network model can be created from existing material flow data with manageable effort. The prediction method is evaluated in detail with multiple data sets and tested against baseline methods from machine learning. The code used for network modeling and evaluation is freely available under a creative commons license. Abstract: Modeling complex systems as networks of interacting elements has gained increased attention in recent years. So far, network modeling in manufacturing and logistics has often focused on the description of system properties. In the data-driven world of smart manufacturing, creating material flow network models becomes a lot easier due to the ubiquitous availability of shop floor and transportation data. At the same time, these highly flexible and continuously changing smart manufacturing systems become less predictive and thus less controllable. This article investigates how the stochastic block model (SBM), a network model with a stochastic description of interconnections, can be applied to model and predict material flows in manufacturing systems. We show how to utilize its properties to forecast the dynamic development of the structure of such systems. The complete process from network modeling using material flow data to the prediction of the future development of the network is demonstrated. Different SBM variants are tested using six company data sets and evaluated in competition with classical machine learning methods for prediction. Our results show that selected SBM variants achieve the best performance in prediction in most scenarios and thus have the potential to play an important role in the management of future dynamic manufacturing systems.