| Popis: |
By leveraging the high dimensional nonlinear mapping capabilities of artificial neural networks in conjunction with the powerful control mechanisms of reinforcement learning, we attain real-time, precise modulation of synthetic jet flow rates over elliptical cylinders with varying aspect ratios (Ar). The training outcomes for elliptical cylinders with a blockage ratio of 0.24 show that for Ar=1 and 0.75, the reward function gradually increases with decreasing oscillations before stabilizing. The agent's control strategy achieved drag reduction rates of 8% and 15%, while 99% of the lift coefficient is effectively suppressed. It deserves emphasis that vortex shedding is entirely eliminated with only 0.1% and 1% of the inlet flow rate. As Ar decreases, the reinforcement learning process slows and becomes less stable, with energy consumption surging to 14.5%, while lift and drag coefficients continue oscillating and vortex shedding remains uncontrolled. When blockage ratio is reduced to 0.12, the reinforcement learning training demonstrates robust convergence and consistent full suppression of vortex shedding across all Ar from 1 to 0.1. Furthermore, drag reduction rates are observed within the range of 6.1% to 32.3%, while the lift coefficient is effectively regulated to remain at zero. For cylinders with Ar between 1 and 0.25, external energy expenditure remains below 1.4% of the inlet flow rate, signifying the realization of both efficient and energy-conservative control strategies within this range. However, for the extremely slender elliptical cylinder with Ar=0.1, the energy cost escalates to 8.1%, underscoring the significantly higher energy expenditure required to fulfill the control objectives for such highly elongated geometries. |
