Multi-objective Optimization Framework for Energy Efficiency and Production Scheduling in Smart Manufacturing Using Reinforcement Learning and Digital Twin Technology Integration

Using Reinforcement Learning and Digital Twin Technology IntegrationManufacturing facilities face concurrent challenges of maximizing production efficiency while reducing energy consumption and environmental impact.Traditional scheduling approaches typically optimize for either production or energy metrics independently, creating a fragmented optimization landscape.This research develops and validates a multi-objective optimization framework integrating reinforcement learning with digital twin technology to simultaneously balance production efficiency and energy consumption in smart manufacturing environments.The research implemented detailed digital twins of three manufacturing facilities in Uzbekistan using Siemens Tecnomatix, integrating real-time data from 387 IoT sensors.A custom-developed deep reinforcement learning algorithm utilizing Proximal Policy Optimization was trained on 18 months of historical data.The framework employed weighted multi-objective functions balancing production, energy, and quality metrics, with validation through A/B testing across 93 production runs.Implementation achieved 22.7% reduction in energy consumption while maintaining production output within 1.2% of baseline capacity.Peak power demand decreased by 27.9%, reducing energy costs by 19.1%.Product quality metrics improved by 6.9% due to optimized machine utilization.The reinforcement learning algorithm demonstrated 89.8% accuracy in predicting energy consumption patterns and achieved convergence 76% faster than conventional optimization approaches.The integrated digital twin-reinforcement learning approach effectively balances energy efficiency and production requirements, creating pathways for sustainable manufacturing without compromising operational performance.

Перевод пока недоступен