A cross-method optimisation framework for automated cheese manufacturing based on digital twin-driven reinforcement learning
International Journal of Management and Decision Making, vol.25, no.5, pp.526-554, 2026 (Scopus)
- Publication Type: Article / Article
- Volume: 25 Issue: 5
- Publication Date: 2026
- Doi Number: 10.1504/ijmdm.2026.156298
- Journal Name: International Journal of Management and Decision Making
- Journal Indexes: Scopus, ABI/INFORM
- Page Numbers: pp.526-554
- Keywords: digital twin, DT, grammage accuracy, heuristic algorithms, industrial cheese manufacturing, reinforcement learning, temperature control optimisation
- Anadolu University Affiliated: Yes
Abstract
This work presents integrated modelling, control synthesis, and simulation for automated production of semi-hard and hard cheeses, focusing on temperature control and grammage accuracy in kaşar cheese production. To overcome the limitations of traditional rule-based and PID control in nonlinear thermal-mechanical environments, eight control and optimisation methods – PID, model predictive control (MPC), genetic algorithms (GA), particle swarm optimisation (PSO), hybrid GA-PSO, reinforcement learning (RL), and a digital twin-coupled MPC/RL approach – are evaluated in an industrial case study under nominal, high-load, and disturbance scenarios. The results show that AI-based approaches outperform traditional controllers by reducing temperature and grammage errors while improving energy efficiency and robustness. These findings demonstrate the effectiveness of combining model-based control, evolutionary optimisation, and learning-based intelligence to achieve robust, energy-efficient, and intelligent cheese manufacturing.