ADAPTIVE METAMODELING: An Application in the Textile Industry

Simulation, a pillar of Industry 4.0, is a powerful tool used to evaluate systems under various conditions, facilitating performance analysis and decision-making. However, it can be time-consuming, especially when optimizing complex models. Metamodeling has emerged as a popular technique for simulation optimization. This paper presents AMSO (Adaptive Metamodeling-based Simulation Optimization), an innovative framework that aims for better solutions with fewer experiments. The approach combines machine learning and metaheuristic techniques to identify and efficiently explore the most promising areas of the solution space. The framework was evaluated in a real-world resource allocation problem of a manufacturing digital twin model. Compared to the Genetic Algorithm, AMSO found a statistically equal solution but required 82% less computational time.