Search

This paper applies deep reinforcement learning (DRL) on the synthetic jet control of flows over an NACA (National Advisory Committee for Aeronautics) 0012 airfoil under weak turbulent condition. Based on the proximal policy optimization method, the appropriate strategy for controlling the mass rate of a synthetic jet is successfully obtained at [Formula: see text]. The effectiveness of the DRL based active flow control (AFC) method is first demonstrated by studying the problem with constant inlet velocity, where a remarkable drag reduction of 27.0% and lift enhancement of 27.7% are achieved, accompanied by an elimination of vortex shedding. Then, the complexity of the problem is increased by changing the inlet velocity condition and reward function of the DRL algorithm. In particular, the inlet velocity conditions pulsating at two different frequencies and their combination are further applied, where the airfoil wake becomes more difficult to suppress dynamically and precisely; and the reward function additionally contains the goal of saving the energy consumed by the synergetic jets. After training, the DRL agent still has the ability to find a proper control strategy, where significant drag reduction and lift stabilization are achieved, and the agent with considerable energy saving is able to save the energy consumption of the synergetic jets for 83%. The performance of the DRL based AFC proves the strong ability of DRL to deal with fluid dynamics problems usually showing high nonlinearity and also serves to encourage further investigations on DRL based AFC.

Based on the deep reinforcement learning (DRL) method, the active flow control strategy obtained from artificial neural networks (ANNs) is applied to reducing the drag force of various blunt bodies. The control strategy is realized by the agent described by ANNs model which maps appropriate environment sensing signals and control actions, and ANNs are constructed by exploring the controlled system through proximal policy optimization method. The drag reduction effect for ellipse, square, hexagon and diamond geometries under double- and triple-jets control is systematically studied, and the robustness of DRL jet control method is verified. The numerical results show that the drag reduction effect of triple-jets control is significantly better than that of double-jets control when Reynolds number is 80 and angle of attack is 0, and under the triple-jets control situation, the DRL agent can significantly reduce the drag by approximately 11.50%, 10.56%, 8.35%, and 2.78% for ellipse, square, hexagon and diamond model, respectively. In addition, based on the ellipse model, the drag reduction effect of the active control strategy under different AOA and different Reynolds numbers are further studied. When the AOA of ellipse configuration are 5°, 10°, 15° and 20° and the Reynolds number remains 80, the control strategies of DRL achieve the drag reduction of 5.44%, 0.59%, 11.67% and 0.28%, respectively. Meanwhile, when the AOA is 0, the drag reduction reaches 10.84% and 23.63% under the condition of the Reynolds number is 160 and 320, respectively. The significant control effect shows that the reinforcement learning method coupled with the ANNs shows a powerful ability to identical system when facing control problem with high-dimensional nonlinear characteristics. The ability to identify complex systems also shows that DRL methods can be further applied to active flow control under conditions of higher Reynolds number.

In the current paper, the active flow control for enhancing vortex induced vibration (VIV) of a circular cylinder, which can be potentially applied in ocean energy harvesting, is achieved by an artificial neural network (ANN) trained through deep reinforcement learning (DRL). The flow past a circular cylinder with and without jet control located on the cylinder is numerically investigated using OpenFOAM, and the ANN is applied to learn an active flow control strategy through experimenting with different mass flow rates of the jets. According to our results, the jets on the cylinder are able to dramatically destabilize the periodic shedding of the cylinder wake, which leads to a much larger VIV and work capability of the cylinder. Through controlling the flow rate of the jets based on the observation of the instantaneous flow field, the ANN successfully increases the drag by 30.78%, and the magnitude of the fluctuation of the drag and lift coefficient by 785.71% and 139.62%, respectively, while the energy consumption of the jets is almost negligible. Furthermore, the net energy output by VIV with jet control increases by 357.63% (case of water) compared with the uncontrolled situation. The results demonstrate that the performance of the active jet control strategy established by DRL for enhancing VIV is outstanding and promising for realizing the transformation from the ocean energy to electrical energy. Therefore, it is encouraged to perform further investigations on VIV enhancement using active flow control based on DRL.

The high-temperature mechanical properties and microstructure of HK40 heat-resistant steel with different content of Al were investigated. The results from scanning electron microscope and transmission electron microscope showed that a large amount of spheroidal and dispersed γ′ phase were observed HK40 steel with 4.72wt.% and 7.10wt.% Al. The diameter of γ′ phase decreases from about 1.5μm to 50nm after solution treatment of 1200°C for 5h. The results of short term tensile test showed that tensile strength at 900°C decreased and the elongation was improved with increasing Al content. The oxides in the alloy with 4.72wt.% and 7.10wt.% Al were more uniform and finer than that in the alloy with and without 1.68wt.% Al.

The aim of this study was to explore the mechanism of the nervous system lesions induced by formaldehyde (FA). Male Balb/c mice were exposed to gaseous formaldehyde for 7 days (8 h/d) with three different concentrations (0, 0.5 and 3.0 mg/m(3)). A group of animals injected with the nitric oxide synthase inhibitor L-NMMA (0.01 mL/g) was also set and exposed to 3.0 mg/m(3) FA. The concentrations of cAMP, cGMP, NO and the activity of NOS in cerebral cortex, hippocampus and brain stem were determined by corresponding assay kits. The results showed that, compared with the control (0 mg/m(3) FA) group, the cAMP contents in cerebral cortex and brain stem were significantly increased in 0.5 mg/m(3) FA group (P0.05), but decreased in 3.0 mg/m(3) FA group (P0.05); The concentration of cAMP in hippocampus was significantly decreased in 3.0 mg/m(3) FA group (P0.05). In comparison with the control group, L-NMMA group showed unchanged cAMP contents and NOS activities in different brain regions, but showed increased cGMP contents in hippocampus and NO contents in cerebral cortex (P0.05). In addition, compared with 3.0 mg/m(3) FA group, L-NMMA group showed increased contents of cAMP and reduced NOS activities in different brain regions, as well as significantly decreased cGMP contents in cerebral cortex and brain stem and NO content in brain stem. These results suggest that the toxicity of FA on mouse nervous system is related to NO/cGMP and cAMP signaling pathways.