Physics–Data Synergy: A Hybrid CFD–Machine Learning Framework for Smart Cold Storage Systems

Abstract

This research presents a groundbreaking physics-data synergy framework that integrates thermal stratification principles with advanced machine learning algorithms to develop intelligent cold storage systems. The study addresses fundamental challenges in thermal management by leveraging the physics of temperature stratification phenomena combined with data-driven optimization methodologies. The physics-based foundation employs validated CFD models demonstrating temperature stratification behavior with thermocline formation occurring across temperature ranges from 275.6K to 363.1K (2.5°C to 90°C). Comprehensive transient analysis reveals distinct thermal evolution patterns through five-stage temperature distribution development, showing progressive stratification establishment over operational time periods. Advanced parameter analysis demonstrates critical relationships between mixing coefficients and inlet velocities, with values ranging from 50,000 to 2,500,000 across different temperature conditions (20°C, 50°C, and 90°C). The data-driven component achieves exceptional predictive accuracy with R² values of 0.96 for temperature stratification prediction and 0.93 for thermal mixing coefficient forecasting. The synergistic framework delivers remarkable performance improvements including 29% reduction in thermal mixing, 38% enhancement in stratification efficiency, and 34% improvement in energy utilization effectiveness. Smart control algorithms developed through physics-data integration enable real-time optimization of thermal gradients, inlet velocity management, and stratification maintenance across multiple operational zones. The framework successfully demonstrates scalability from laboratory-scale thermal storage systems to industrial cold storage applications while maintaining high performance standards essential for commercial deployment.