Browsing by Author "Zhang, Xin-Lei"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Acoustic Inversion for Uncertainty Reduction in Reynolds-Averaged Navier-Stokes-Based Jet Noise Prediction
    Zhang, Xin-Lei; Xiao, Heng; Wu, Ting; He, Guowei (American Institute of Aeronautics and Astronautics, 2021-12-13)
    The Reynolds-averaged Navier–Stokes (RANS)-based method is a practical tool to provide rapid assessment of jet noise-reduction concepts. However, the RANS-based method requires modeling assumptions to represent noise generation and propagation, which often reduces the predictive accuracy due to the model-form uncertainties. In this work, the ensemble Kalman filter-based acoustic inversion method is introduced to reduce uncertainties in the turbulent kinetic energy and dissipation rate based on the far-field noise and the axial centerline velocity data. The results show that jet noise data are more effective from which to infer turbulent kinetic energy and dissipation rate compared to velocity data. Moreover, the inferred noise source is able to improve the estimation of the turbulent flowfield and the far-field noise at unobserved locations. Further, the noise model parameters are also considered uncertain quantities, demonstrating the ability of the proposed framework to reduce uncertainties in both the RANS and noise models. Finally, one realistic case with experimental data is investigated to show the practicality of the proposed framework. The method opens up the possibility for the inverse modeling of jet noise sources by incorporating far-field noise data that are relatively straightforward to be measured compared to the velocity field.
  • Thumbnail Image
    Ensemble Gradient for Learning Turbulence Models from Indirect Observations
    Strofer, Carlos A. Michelen; Zhang, Xin-Lei; Xiao, Heng (Global Science Press, 2021-11-01)
    Training data-driven turbulence models with high fidelity Reynolds stress can be impractical and recently such models have been trained with velocity and pressure measurements. For gradient-based optimization, such as training deep learning models, this requires evaluating the sensitivities of the RANS equations. This paper explores the use of an ensemble approximation of the sensitivities of the RANS equations in training data-driven turbulence models with indirect observations. A deep neural network representing the turbulence model is trained using the network’s gradients obtained by backpropagation and the ensemble approximation of the RANS sensitivities. Different ensemble approximations are explored and a method based on explicit projection onto the sample space is presented. As validation, the gradient approximations from the different methods are compared to that from the continuous adjoint equations. The ensemble approximation is then used to learn different turbulence models from velocity observations. In all cases, the learned model predicts improved velocities. However, it was observed that once the sensitivity of the velocity to the underlying model becomes small, the approximate nature of the ensemble gradient hinders further optimization of the underlying model. The benefits and limitations of the ensemble gradient approximation are discussed, in particular as compared to the adjoint equations.