Simulation-based metaheuristic optimization algorithm for material handling

Modern technologies and the emergent Industry 4.0 paradigm have empowered the emergence of flexible production systems suitable to cope with custom product demands, typical in this era of competitive marketplaces. However, production flexibility claims periodic changes in the setup of production facilities. The level of flexibility of a production process increases as the reconfiguration capacity of its facilities increases. Nevertheless, doing that efficiently requires accurate coordination between productive resources, task planning, and decision-making systems aiming to maximize value for the client, minimizing non-added-value production tasks, and continuous process improvement. In a manufacturing system, material handling within manufacturing facilities is one of the major non-value-added tasks strongly affected by changes in plant floor layouts and demands for producing customized products. This work proposes a metaheuristic simulation-based optimization methodology to address the material handling problem in dynamic environments. Our proposed approach integrates optimization, discrete event simulation, and artificial intelligence methods. Our proposed optimization algorithm is mainly based on the ideas of the novel population-based optimization algorithm called Q-learning embedded Sine Cosine Algorithm, inspired by the Sine Cosine Algorithm. Unlike those, our proposed approach can deal with discrete optimization problems. It includes in its formulation a reinforcement learning embedded algorithm for the self-learning of the parameters of the metaheuristic optimization algorithm, and discrete event simulation is used for simulating the shop floor operations. The performance of the proposed approach is evaluated through an exhaustive analysis of simple to complex cases. In addition, a comparison is made with other comparable optimization methodologies.