Obtaining data-based decision support faster is a competitive advantage for companies. Faster decisions means that companies can capitalize on opportunities and avert costly mistakes. Simulation as a predictive tool, with the rise of digitalization, is used across many disciplines in the manufacturing industry. When analyzing current and future production lines, more companies are using discrete event simulation software which offers improved results and accuracy compared to static analysis tools. If simulation is coupled with multi-objective optimization and knowledge extraction, new possibilities for production systems are introduced where artificial intelligence can be used to improve the systems.
Seeking predictive answers on the manufacturing network level by re-using line models is problematic. Simulation models on the line level are usually detailed to answer specific questions about that production line. Trying to connect several of these models with the intent to optimize the complete manufacturing network will create computationally expensive models. Reducing the complexity of the line models is not enough, a new representation of the line models is needed. New methods for the aggregation of detailed line model data into a faster and more computationally efficient line modules will enable analysis and optimization of manufacturing networks.
A novel method for the aggregation of detailed simulation model data to a more efficient meta-model is one part of the expected result for this project. Increasing the generalizability of the method is critical and application studies will be performed on different types of manufacturing processes. After the generalizability has been verified, extending the method to also incorporate the possibility of optimization and knowledge extraction, together with which input data is required, a framework can be created. This aggregation framework will be the main contribution from this project.
Research proposal, PhD programme, University of Skövde