Optimizing Thermal Control in Pulsating Jets through Deep Reinforcement Learning
Please login to view abstract download link
This research study focuses on Deep Reinforcement Learning (DRL) utilization for thermal control based on Computational Fluid Dynamics (CFD). The study focuses on forced convection on a hot plate subjected to a pulsating cooling jet with a variable velocity. First, the applicability of a classical Deep Q-Network (DQN) has been assessed. Subsequently, a comprehensive analysis of DRL variants namely Soft and Hard Double and Duel DQN has been performed. Results highlight the superiority of Soft Double and Duel DQN in maintaining the temperature within the desired threshold for over 98% of the control cycle. These findings underscore the promising potential of DRL in addressing thermal control systems.
