Challenges in Controlling Jet Entrainment and Diffusion by Initial Spatial Velocity Distribution

Abstract

Jet flows are phenomena observed in various engineering applications, and proposing control methods and techniques for jets is crucial for the realization of high-performance fluidic devices. In jet flow control, manipulating the initial velocity distribution is the most common approach. The parameters considered here include the magnitude of the velocity and the ejection angle. From the perspective of engineering feasibility, it is essential to clarify the impact and importance of optimizing the ejection angle components. Therefore, in this study, we investigated the effects of optimizing the ejection angle components of the initial velocity using deep reinforcement learning combined with fluid dynamics simulations. As a result, it was found that optimization including the ejection angle components significantly improves performance when the control objective is related to entrainment suppression. On the other hand, when the control objective is related to entrainment promotion, the ejection angle components were found to have little impact on the optimization.