-
An extension of the compound flow theory with shear and friction
Authors:
Jan Van den Berghe,
Miguel Alfonso Mendez,
Yann Bartosiewicz
Abstract:
Compound flows consist of two or more parallel compressible streams inside a duct. Its theoretical treatment has recently been applied to model ejectors, but also to post-process and to interpret CFD results. More precisely, it has been observed that compound flows can choke upstream of the geometrical throat. While it is widely known that friction can push the sonic section downstream, no mechani…
▽ More
Compound flows consist of two or more parallel compressible streams inside a duct. Its theoretical treatment has recently been applied to model ejectors, but also to post-process and to interpret CFD results. More precisely, it has been observed that compound flows can choke upstream of the geometrical throat. While it is widely known that friction can push the sonic section downstream, no mechanism has been identified to explain the movement in the opposite direction.
In this work, we extend the existing compound flow theory with shear and friction. The resulting model captures the up- and downstream movement of the sonic section. We show analytically that shear between the streams is the driving force behind the upstream movement by analysing the singularity of the governing equations in the sonic section. Additionally, we provide a quasi-third order approximation of the governing equations in the vicinity of the sonic section to alleviate potential numerical issues. Finally, we compare the predictions of the model with post-processed CFD results of a compound nozzle. We study the individual effects of shear and friction through an inviscid simulation for the isentropic case and a viscid simulation with a slip or no-slip condition at the wall to include shear with or without friction.
The proposed extension offers new possibilities in modelling and analysis of compound flows.
△ Less
Submitted 14 January, 2024;
originally announced January 2024.
-
Reinforcement Twinning: from digital twins to model-based reinforcement learning
Authors:
Lorenzo Schena,
Pedro Marques,
Romain Poletti,
Samuel Ahizi,
Jan Van den Berghe,
Miguel A. Mendez
Abstract:
Digital twins promise to revolutionize engineering by offering new avenues for optimization, control, and predictive maintenance. We propose a novel framework for simultaneously training the digital twin of an engineering system and an associated control agent. The twin's training combines adjoint-based data assimilation and system identification methods, while the control agent's training merges…
▽ More
Digital twins promise to revolutionize engineering by offering new avenues for optimization, control, and predictive maintenance. We propose a novel framework for simultaneously training the digital twin of an engineering system and an associated control agent. The twin's training combines adjoint-based data assimilation and system identification methods, while the control agent's training merges model-based optimal control with model-free reinforcement learning. The control agent evolves along two independent paths: one driven by model-based optimal control and the other by reinforcement learning. The digital twin serves as a virtual environment for confrontation and indirect interaction, functioning as an "expert demonstrator." The best policy is selected for real-world interaction and cloned to the other path if training stagnates. We call this framework Reinforcement Twinning (RT). The framework is tested on three diverse engineering systems and control tasks: (1) controlling a wind turbine under varying wind speeds, (2) trajectory control of flap**-wing micro air vehicles (FWMAVs) facing wind gusts, and (3) mitigating thermal loads in managing cryogenic storage tanks. These test cases use simplified models with known ground truth closure laws. Results show that the adjoint-based digital twin training is highly sample-efficient, completing within a few iterations. For the control agent training, both model-based and model-free approaches benefit from their complementary learning experiences. The promising results pave the way for implementing the RT framework on real systems.
△ Less
Submitted 11 July, 2024; v1 submitted 6 November, 2023;
originally announced November 2023.
-
A 1D Model for the Unsteady Gas Dynamics of Ejectors
Authors:
Jan Van den Berghe,
Bruno R. B. Dias,
Yann Bartosiewicz,
Miguel A. Mendez
Abstract:
We propose a 1D unsteady model for supersonic single-phase ejectors. The model treats an ejector as a pipe network with two inputs and one output and combines a 1D gas dynamics formulation in each `pipe' with a junction model for entrainment and mixing. The model is calibrated and validated on experimental data in steady-state conditions and used to analyze the choking mechanism for the mixed flow…
▽ More
We propose a 1D unsteady model for supersonic single-phase ejectors. The model treats an ejector as a pipe network with two inputs and one output and combines a 1D gas dynamics formulation in each `pipe' with a junction model for entrainment and mixing. The model is calibrated and validated on experimental data in steady-state conditions and used to analyze the choking mechanism for the mixed flow. The model was then benchmarked against 2D URANS simulations to predict the ejector response to a sudden change in operating conditions, producing traveling waves. The results show that the model can correctly predict the ejector performance and the stream-wise evolution of relevant integral quantities (e.g. mass flow rates and momentum) in both steady and transient conditions.
△ Less
Submitted 16 August, 2022;
originally announced August 2022.
-
Challenges and Opportunities for Machine Learning in Fluid Mechanics
Authors:
M. A. Mendez,
J. Dominique,
M. Fiore,
F. Pino,
P. Sperotto,
J. Van den Berghe
Abstract:
Big data and machine learning are driving comprehensive economic and social transformations and are rapidly re-sha** the toolbox and the methodologies of applied scientists. Machine learning tools are designed to learn functions from data with little to no need of prior knowledge. As continuous developments in experimental and numerical methods improve our ability to collect high-quality data, m…
▽ More
Big data and machine learning are driving comprehensive economic and social transformations and are rapidly re-sha** the toolbox and the methodologies of applied scientists. Machine learning tools are designed to learn functions from data with little to no need of prior knowledge. As continuous developments in experimental and numerical methods improve our ability to collect high-quality data, machine learning tools become increasingly viable and promising also in disciplines rooted in physical principles. These notes explore how machine learning can be integrated and combined with more classic methods in fluid dynamics. After a brief review of the machine learning landscape, we show how many problems in fluid mechanics can be framed as machine learning problems and we explore challenges and opportunities. We consider several relevant applications: aeroacoustic noise prediction, turbulence modelling, reduced-order modelling and forecasting, meshless integration of (partial) differential equations, super-resolution and flow control. While this list is by no means exhaustive, the presentation will provide enough concrete examples to offer perspectives on how machine learning might impact the way we do research and learn from data.
△ Less
Submitted 13 April, 2024; v1 submitted 25 February, 2022;
originally announced February 2022.
-
Artificial Neural Networks Modelling of Wall Pressure Spectra Beneath Turbulent Boundary Layers
Authors:
J. Dominique,
J. Van den Berghe,
C. Schram,
M. A. Mendez
Abstract:
We analyse and compare various empirical models of wall pressure spectra beneath turbulent boundary layers and propose an alternative machine learning approach using Artificial Neural Networks (ANN). The analysis and the training of the ANN are performed on data from experiments and high-fidelity simulations by various authors, covering a wide range of flow conditions. We present a methodology to…
▽ More
We analyse and compare various empirical models of wall pressure spectra beneath turbulent boundary layers and propose an alternative machine learning approach using Artificial Neural Networks (ANN). The analysis and the training of the ANN are performed on data from experiments and high-fidelity simulations by various authors, covering a wide range of flow conditions. We present a methodology to extract all the turbulent boundary layer parameters required by these models, also considering flows experiencing strong adverse pressure gradients. Moreover, the database is explored to unveil important dependencies within the boundary layer parameters and to propose a possible set of features from which the ANN should predict the wall pressure spectra. The results show that the ANN outperforms traditional models in adverse pressure gradients, and its predictive capabilities generalise better over the range of investigated conditions. The analysis is completed with a deep ensemble approach for quantifying the uncertainties in the model prediction and integrated gradient analysis of the model sensitivity to its inputs. Uncertainties and sensitivities allow for identifying the regions where new training data would be most beneficial to the model's accuracy, thus opening the path towards a self-calibrating modelling approach.
△ Less
Submitted 10 January, 2022;
originally announced January 2022.
