Optimizing simulated manufacturing systems using machine learning coupled to evolutionary algorithms

Recent works have shown that simulation optimization of manufacturing systems can be efficiently addressed using evolutionary algorithms. The main drawbacks of these algorithms are that they are notoriously slow and that they bring no understanding on the behavior of the system. So we propose to add to these algorithms a machine learning module, which can highlights several critical parameters and guide then the research of solution. The benefits of this approach are demonstrated through the example of optimizing an assembly kanban system.