Application of Digital Twin and simulation for dynamic task allocation

The topic covered in this paper explains a practical application of the project created in the Mind4Lab laboratory of the Politecnico di Torino, with the aim of implementing a Digital Twin, an extremely promising technology in the manufacturing and other fields, research on which is still ongoing. The assumption that was made at the beginning of the project was that, since the DT is able to replicate the behavior of a system in real time, then by exploiting this data it would be possible to work in a dynamic environment and making decisions consequently to the changes detected in the physical system. The use of simulation models as a decision-making tool at a company level is a great strategy to solve industrial problems, especially in the manufacturing field, which is why it was chosen as the environment to which to apply the case study. The reasons for this are related to the wide variety of problems that can arise within a production line, to name a few it is possible to find machinery maintenance, staff absences, and unforeseen changes in scheduling. The decision of the specific field to be analyzed was made by looking for the one that would allow the case study to be developed as broadly and comprehensively as possible. The staff absence scenario was ruled out regardless, as a virtual model of human operators would have been extremely error-prone due firstly to the variability of system behavior, which is impossible to predict, and secondly to the practical and technical difficulty of collecting data from such resources. The two candidates from which the decision was made were therefore the topic of predictive maintenance and the topic of dynamic scheduling. Both gave the possibility to monitor a physical production system through sensors, and to collect and analyze data in order to evaluate performance; the reason why the topic chosen in the end was the dynamic scheduling is related to the fact that it also gives the possibility to see how machinery interacts with each other when a variation in production compromises the planned scheduling, thus adding a focus on the connection within the physical system, and not just the connections between physical and virtual and vice versa. Dynamic scheduling is an evolution of the Predictive scheduling. The latter is a strategy widely used for the preparation of a good-quality optimized baseline schedule, which is done in advance taking into account many possible scenarios, and easy to maintain. Although it turns out to be a good technique, it is important to remind that in the manufacturing context, the integrity of a schedule is very easy to be disrupted. The effect of previously introduced events may in some cases lead to the need to completely reschedule the project; Dynamic scheduling is useful in that it allows a strategy on how to first create the initial baseline, but more importantly, in times of need, a strategy on how to respond to events in real time. The hypothesis to be proved is that, following a generic production line failure, the final results in a dynamic environment should perform better than the theoretical results calculated a priori. The entire work will therefore be carried out with the aim of collecting data in favor of this thesis, and consequently demonstrating that the application of a digital twin is able to bring benefits in terms of performance.