Stochastic Complex Networks as Predictive and Explanatory Model for the Dynamic Development of Production Logistic Systems

• Applicant: Professor Dr. Till Becker
• Subject Area: Production Systems, Operations Management, Quality Management and Factory Planning
• Term: from 2016 to 2020
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 310784388

Production logistic systems are composed of the physical resources of the manufacturing system, but also of raw material, products, processes, orders, plans, etc., which are required to complete the value creation process. Job shop environments in particular show complex structures and dynamic behavior, so that the anticipation of changes within the system is complicated. These structural changes include, e.g., the necessity to introduce new machines, the placement of machines on the shop floor, the decommissioning of machines, the installation of new transportation routes, the close-down of obsolete transportation routes, etc. At the same time, there is a rising need for controllability and predictability of changes in production logistic systems. This is caused by the development towards shorter product life cycles and a higher amount of variants on the one hand, in combination with increasing cost pressure due to the globalization on the other hand. If companies are not successful in the timely adaption of their structures in manufacturing, they will face competitive disadvantages.The goal of this project is to create reliable forecasts of structural changes in a manufacturing system with a stochastic model of the material flow with comparably low effort. The basic assumption is that there are predominant patterns in material flow networks, which are more probable to observe in comparison to other patterns. The approach is to model the material flow in a job shop as a complex network and to create a so called Stochastic Block Model (SBM) based on the network model. This SBM serves as a prediction model for various types of changes in a network representation of the manufacturing system. Real material flow data from the IT systems of manufacturers serve as input for the model creation. The quality of the prognosis will be compared to the prognosis results of state-of-the-art machine learning approaches using the same data.The result of the project is the concept and the evaluation of a new, effortless approach in the field of manufacturing systems for control of dynamic and complex production logistic systems by the prognosis of structural changes. The project offers the opportunity to extend the application of the approach in a subsequent project to a broader field, such as logistic processes in general.

• DFG Programme: Research Grants