• Corpus ID: 237562959

Assessments of model-form uncertainty using Gaussian stochastic weight averaging for fluid-flow regression

  title={Assessments of model-form uncertainty using Gaussian stochastic weight averaging for fluid-flow regression},
  author={Masaki Morimoto and Kai Fukami and Romit Maulik and Ricardo Vinuesa and Koji Fukagata},
We use Gaussian stochastic weight averaging (SWAG) to assess the model-form uncertainty associated with neural-network-based function approximation relevant to fluid flows. SWAG approximates a posterior Gaussian distribution of each weight, given training data, and a constant learning rate. Having access to this distribution, it is able to create multiple models with various combinations of sampled weights, which can be used to obtain ensemble predictions. The average of such an ensemble can be… 


Quantifying and reducing model-form uncertainties in Reynolds-averaged Navier-Stokes simulations: A data-driven, physics-informed Bayesian approach
An open-box, physics-informed Bayesian framework for quantifying model-form uncertainties in RANS simulations and results suggest that, even with very sparse observations, the posterior mean velocities and other QoIs have significantly better agreement with the benchmark data compared to the baseline results.
Predictions of turbulent shear flows using deep neural networks
In the present work we assess the capabilities of neural networks to predict temporally evolving turbulent flows. In particular, we use the nine-equation shear flow model by Moehlis et al. [New J.
Convolutional neural network for transition modeling based on linear stability theory
A novel convolutional neural network inspired by the underlying physics as described by the stability equations is proposed, and it is demonstrated that, for Tollmien-Schlichting instabilities in two-dimensional, low-speed boundary layers, the proposed network encodes information in the boundary layer profiles into an integral quantity that is strongly correlated to a well-known, physically defined parameter -- the shape factor.
Quantification of Uncertainties in Turbulence Modeling: A Comparison of Physics-Based and Random Matrix Theoretic Approaches
Abstract Numerical models based on Reynolds-Averaged Navier-Stokes (RANS) equations are widely used in engineering turbulence modeling. However, the RANS predictions have large model-form
Model Order Reduction with Neural Networks: Application to Laminar and Turbulent Flows
The capability of neural network-based model order reduction, i.e., autoencoder (AE), for fluid flows, which comprises of convolutional neural networks and multi-layer perceptrons is investigated, finds that the AE models are sensitive to the choice of the aforementioned parameters depending on the target flows.
A Bayesian Calibration–Prediction Method for Reducing Model-Form Uncertainties with Application in RANS Simulations
The extended framework is a Bayesian calibration–prediction method for reducing model-form uncertainties in RANS simulations for flows where there are no available data on the flow to be predicted, and the merits of the proposed method are demonstrated on two flows that are challenging to standard RANS models.
Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
A new theoretical framework is developed casting dropout training in deep neural networks (NNs) as approximate Bayesian inference in deep Gaussian processes, which mitigates the problem of representing uncertainty in deep learning without sacrificing either computational complexity or test accuracy.
Sparse identification of nonlinear dynamics with low-dimensionalized flow representations
The proposed scheme with an appropriate parameter choice enables the proposed scheme to analyze high-dimensional nonlinear dynamics with interpretable low-dimensional manifolds to examine the applicability of SINDy to turbulence with low- dimensional modes.
A deep learning enabler for nonintrusive reduced order modeling of fluid flows
A modular deep neural network (DNN) framework for data-driven reduced order modeling of dynamical systems relevant to fluid flows is introduced and can be used as a robust predictive tool for non-intrusive model order reduction of complex fluid flows.
Turbulence Modeling in the Age of Data
Recent developments in bounding uncertainties in RANS models via physical constraints, in adopting statistical inference to characterize model coefficients and estimate discrepancy, and in using machine learning to improve turbulence models are surveyed.