Self-Adaptive Optimization Techniques for Matrix Production Systems

The Industry 4.0 era has enabled the concept of Matrix Production Systems, where products can be manufactured on demand by dynamically reconfiguring independent modules on the shop floor. This reconfiguration process, however, is not trivial, since certain possibilities incur much higher reconfiguration costs than others. To tackle this challenge, researchers have employed optimization techniques for identifying the best possible reconfiguration solutions. However, these approaches do not consider the self-adaptive nature of a manufacturing system, i.e., the need to re-optimize configurations when the order queue changes. To address this limitation, this paper focuses on the self-adaptive reconfiguration of Matrix Production System modules. Specifically, it proposes a genetic algorithm-based approach for self-adaptive reconfiguration, introducing the use of a solution archive mechanism. This approach is compared to an integer linear programming model, using a dataset that features dynamic updates to the queue of orders. Results show that the genetic algorithm-based approach generally has lower execution times than integer linear programming, though it does not converge to the global optimum.