Results in Engineering (Mar 2025)
Optimizing Rotary Cement Kiln modelling: A comparative analysis of metaheuristics in a real-world application
Abstract
This study introduces a real-world metaheuristic application designed to minimize the simulation errors of a novel one-dimensional (1D) numerical model of a rotary cement kiln, calibrated by Computational Fluid Dynamics (CFD) results. This model aims to significantly reduce simulation time, facilitating the analysis of multiple scenarios. A major challenge is the high proportion of invalid or infeasible solutions typical in industrial environments, compounded by the time-sensitive nature of real-world applications. Another notable difficulty lies in selecting among the plethora of metaheuristic methods, which may exhibit significant variations in search efficiency. To address these challenges, this work evaluates the performance of five state-of-the-art metaheuristics and examines the impact of two penalty methods, death and static, on solution quality, constrained by a limited number of model evaluations due to time constraints. Furthermore, an adjustment to the Marine Predators Algorithm (MPA) is proposed, targeted at enhancing its exploitation phase in alignment with the established stopping criteria. The algorithm design and selection process adhere to recommendations from literature review articles, employing robust statistical techniques, including Bayesian analysis and Monte Carlo sampling to support the replicability of analysis. Results indicate that the adapted MPA is expected to achieve similar error reduction to the other methods but requires 40 % fewer evaluations, thus positioning as the most suitable for the industrial application. The findings are anticipated to provide practical insights for researchers and developers by presenting a case study grounded in best practices to support more effective algorithm design and informed decision-making in real-world optimization challenges.
