1 May 2020 in Physics of Fluids
This paper focuses on the active flow control of a computational fluid dynamics simulation over a range of Reynolds numbers using deep reinforcement learning (DRL). More precisely, the proximal policy optimization (PPO) method is used to control the mass flow rate of four synthetic jets symmetrically located on the upper and lower sides of a cylinder immersed in a two-dimensional flow domain. The learning environment supports four flow configurations with Reynolds numbers 100, 200, 300, and 400, respectively. A new smoothing interpolation function is proposed to help the PPO algorithm learn to set continuous actions, which is of great importance to effectively suppress problematic jumps in lift and allow a better convergence for the training process. It is shown that the DRL controller is able to significantly reduce the lift and drag fluctuations and actively reduce the drag by ∼5.7%, 21.6%, 32.7%, and 38.7%, at Re = 100, 200, 300, and 400, respectively. More importantly, it can also effectively reduce drag for any previously unseen value of the Reynolds number between 60 and 400. This highlights the generalization ability of deep neural networks and is an important milestone toward the development of practical applications of DRL to active flow control.
1 February 2023 in Physics of Fluids
This work introduces a novel data-driven framework to formulate explicit algebraic Reynolds-averaged Navier–Stokes (RANS) turbulence closures. Recent years have witnessed a blossom in applying machine learning (ML) methods to revolutionize the paradigm of turbulence modeling. However, due to the black-box essence of most ML methods, it is currently hard to extract interpretable information and knowledge from data-driven models. To address this critical limitation, this work leverages deep learning with symbolic regression methods to discover hidden governing equations of Reynolds stress models. Specifically, the Reynolds stress tensor is decomposed into linear and non-linear parts. While the linear part is taken as the regular linear eddy viscosity model, a long short-term memory neural network is employed to generate symbolic terms on which tractable mathematical expressions for the non-linear counterpart are built. A novel reinforcement learning algorithm is employed to train the neural network to produce best-fitted symbolic expressions. Within the proposed framework, the Reynolds stress closure is explicitly expressed in algebraic forms, thus allowing for direct functional inference. On the other hand, the Galilean and rotational invariance are craftily respected by constructing the training feature space with independent invariants and tensor basis functions. The performance of the present methodology is validated through numerical simulations of three different canonical flows that deviate in geometrical configurations. The results demonstrate promising accuracy improvements over traditional RANS models, showing the generalization ability of the proposed method. Moreover, with the given explicit model equations, it can be easier to interpret the influence of input features on generated models.
1 May 2023 in Physics of Fluids
The past few years have witnessed a renewed blossoming of data-driven turbulence models. Quantification of the concomitant modeling uncertainty, however, has mostly been omitted, and the generalization performance of the data-driven models is still facing great challenges when predicting complex flows with different flow physics not seen during training. A robust data-driven Reynolds-averaged turbulence model with uncertainty quantification and non-linear correction is proposed in this work with the Bayesian deep neural network. In this model, the Reynolds stress tensor is decomposed into linear and non-linear parts. The linear part is taken as the usual linear eddy viscosity model while the non-linear counterpart is learned by a Bayesian deep neural network. Independent tensor bases of invariants and tensors constituted by mean strain rate tensor and rotation rate tensor are embedded into the neural network to effectively consider key turbulence features in different flows. The proposed model is well validated through numerical simulations of four canonical flows that significantly deviate in geometrical configurations and/or Reynolds numbers from those in the training data. With the non-linear corrections of embedded invariants and tensors representing key features of turbulence, the proposed model not only improves the predictive capabilities of Reynolds-averaged turbulence models on the same mesh but also has better generalization performance when simulating complex turbulent flows with large scale separation. In addition, this model allows us to quantitatively demonstrate the confidence interval of the predicted flow quantities that are originated from the model itself.
1 July 2025 in Physics of Fluids
At the stage of overall shape design for high-speed aircraft, it is challenging to determine how the overall geometric features affect aerodynamic characteristics. Developing a set of geometric characteristic parameters that can effectively represent the aerodynamic characteristics of the aircraft with strong generalization ability is crucial to addressing this issue. Taking the lifting body as the research object, we propose the ratio of the projected area in the Z and Y directions to the theoretical no backward sweeping Z and Y projected area (SZ/SZT, SY/SYT) to characterize the existence of multistage swept-back features among the lifting bodies. In this study, we investigated the effect of parameter SY/SYT on aerodynamic characteristics through a two-dimensional flow model based on shock wave and expansion wave theories. To validate the effectiveness of the proposed parameters, neural networks were employed for aerodynamic modeling of aircraft, with a focused investigation on the effects of parameters SZ/SZT and SY/SYT on network generalization capability and prediction accuracy. A parametric approach was adopted to generate lifting-body configurations, and an aerodynamic database was established through engineering estimation methods. The results demonstrate that the proposed characteristic parameters SZ/SZT and SY/SYT reduce the prediction error by 34% and decrease the training data requirement by approximately one-third while maintaining equivalent error levels. Systematic parameter influence analysis elucidates the mapping relationships between key geometric characteristic parameters of lifting-body configurations and their aerodynamic performance.
1 January 2025 in Physics of Fluids
To overcome the problems of high computational complexity, low efficiency, and limited accuracy in the iterative solution of traditional high-order constitutive equations, this paper proposes a deep learning-based high-order constitutive equation reconstruction (DLHCER) method. The method embeds the nonlinear terms into the source terms of the Navier–Stokes (NS) equations, and the DLHCER control equations are obtained by physical reconstruction. In the training stage, the error between the experimental data and computational fluid dynamics simulation data is utilized to generate the training data and establish the mapping relationship between the free incoming flow variables and the characteristic nonlinear variables. Ultimately, the complex nonlinear terms in DLHCER are fitted and predicted by an agent model to capture the complex relationship between the source term parameters and the incoming flow conditions, thus replacing the traditional iterative solution process. Experimental results show that the average relative error between the theoretical and predicted values of the DLHCER model is 0.302%. The agent model is further analyzed for Shapley additive explanations interpretability to clarify the relationship between network inputs and outputs. Through the comparative validation of nonlinear coupled constitutive relations, Navier–Stokes–Fourier and DLHCER on one-dimensional positive excitation wave arithmetic cases, the results show that DLHCER significantly simplifies the computational process by directly correcting the free parameters of the source term, avoids the cumbersome iterative process, and realizes the double enhancement of computational efficiency and accuracy. Therefore, DLHCER is expected to provide strong support for higher-order constitutive equations in engineering applications in the field of dilute gases.
1 January 2026 in Physics of Fluids
This paper focuses on the reduced-order modeling of strongly compressible flows over a range of parameters based on the combination of proper orthogonal decomposition (POD) and Kolmogorov–Arnold networks (KAN). Precisely, the present method first utilizes POD to extract the dominant modal features of the flow field. Then, the powerful high-dimensional nonlinear mapping capability of KAN is leveraged to reconstruct the complex dynamic behavior of shock waves. The results of several typical hypersonic flows imply that the proposed reduced-order model (ROM) significantly improves computational efficiency while demonstrating higher prediction accuracy than most existing intelligent ROM methods. More importantly, the present ROM can also effectively predict the flow fields for many unseen flow parameters. These results highlights the generalization performance of the proposed method and its remarkable advantages in predicting complex compressible flows.