Traditionally, simulation has been used to evaluate production systems, but didn’t focus on control and observation issues. Traditional simulation-driven experiments focus on the evaluation of the response (lead times, resource utilization, min/average/max queue occupation) of a system under a certain load: for instance, to evaluate the size of an input buffer in front of a machine in a shop floor, or the maximum waiting time of customers in a call center... The model used are inspired by a queuing-theory formalism: they are structured as a set of server and queues, entities arriving into the system following a distribution law, and visiting servers in a pre-determined sequence.
If this modeling practice is well adapted to dimensionment issues, it is not easily usable if it comes to evaluate more precisely manufacturing control issues. Indeed, a big issue when modeling complex manufacturing systems is to model decisions taken by operators in the shop floor. This is partially caused by the fact that queuing-theory simulation component have often an implicit behavior. For instance, when an entity arrive in front of a server, its treatment begin as soon as the server is free. As a result, practitioners must use modeling “tricks” to integrate more complex decisional behavior into the model. Because of this, the comparison of several control strategies require to redevelop the whole model.
Another point that was not easy to study using traditional software is the impact of products observation on manufacturing control. New identification technologies (one of the most popular being RFID) that are becoming available require a tool able to model various observation strategies, in order to compare them. For instance, to access the ROI of a planned RFID system, or to determine the optimal placement of readers.
Finally, Emulica aims at a more realistic modeling of the physical constraints of a shop floor. For instance, traditional simulation models includes queues of entities able to sort entity based on a criteria. Self-sorting queues don’t exist in a real shop floor, even if they may be handful for simulating flows in a macroscopic point of view. By using more realistic models, we will be able to build a virtual -emulated- shop floor, that could be seamlessly replaced by the real one, thus reducing the development effort of the control system.