Smart HVAC Heat Exchanger Network Optimization Through Collaborative IoT-Enabled Predictive Analytics for Manufacturing Facilities

Abstract

Manufacturing facilities rely on sophisticated Heating, Ventilation, and Air Conditioning (HVAC) systems to ensure precise environmental conditions; however, operating these systems in isolated silos often results in substantial energy inefficiencies. This study addresses this challenge by developing and validating a collaborative Internet of Things enabled framework that optimizes heat exchanger networks using privacy-preserving predictive analytics. A distributed IoT architecture comprising 1,234 sensors was deployed across eight diverse manufacturing facilities (chemical, electronics, and automotive) in Saudi Arabia. The framework utilized federated Long Short-Term Memory neural networks. Using the Federated Averaging algorithm, these networks collaboratively trained a global optimization model without sharing proprietary local data. Over a 12-month operational period compared against a three-month baseline, the framework achieved a 29.1% average reduction in HVAC energy consumption (p < 0.001) and improved temperature control precision by 37%. Furthermore, the federated learning model significantly outperformed isolated control strategies, reducing prediction error by 61.8% and preventing 94% of inter-zonal operational conflicts. These results demonstrate that collaborative, privacy-preserving intelligence offers a scalable, robust solution for industrial energy management, effectively bridging the gap between localized control and system-wide optimization in support of Industry 5.0 sustainability goals.

Article history: Received (August 19, 2025); Revised (December 5, 2025); Accepted (January 23, 2026); Published online (April 23, 2026)