-
Analysis and Simulation of a Coupled Fluid-Heat System in a Thin, Rough Layer
Authors:
Tom Freudenberg,
Michael Eden
Abstract:
We investigate the effective coupling between heat and fluid dynamics within a thin fluid layer in contact with a solid structure via a rough surface. Moreover, the opposing vertical surfaces of the thin layer are in relative motion. This setup is particularly relevant to grinding processes, where cooling lubricants interact with the rough surface of a rotating grinding wheel. The resulting model…
▽ More
We investigate the effective coupling between heat and fluid dynamics within a thin fluid layer in contact with a solid structure via a rough surface. Moreover, the opposing vertical surfaces of the thin layer are in relative motion. This setup is particularly relevant to grinding processes, where cooling lubricants interact with the rough surface of a rotating grinding wheel. The resulting model is non-linearly coupled through(i) temperature-dependent viscosity and (ii) convective heat transport. The underlying geometry is highly heterogeneous due to the thin, rough surface characterized by a small parameter representing both the height of the layer and the periodicity of the roughness. We analyze this non-linear system for existence, uniqueness, and energy estimates and study the limit behavior within the framework of two-scale convergence in thin domains. In this limit, we derive an effective interface model in 3D (a line in 2D) for the heat and fluid interactions inside the fluid. We implement the system numerically and validate the limit problem through direct comparison with the micromodel. Additionally, we investigate the influence of the temperature-dependent viscosity and various geometrical configurations via simulation experiments. The corresponding numerical code is freely available on GitHub.
△ Less
Submitted 4 June, 2024;
originally announced June 2024.
-
Homogenization and simulation of heat transfer through a thin grain layer
Authors:
Tom Freudenberg,
Michael Eden
Abstract:
We investigate the effective influence of grain structures on the heat transfer between a fluid and solid domain using mathematical homogenization. The presented model consists of heat equations inside the different domains, coupled through either perfect or imperfect thermal contact. The size and the period of the grains is of order $\varepsilon$, therefore forming a thin layer. The equation para…
▽ More
We investigate the effective influence of grain structures on the heat transfer between a fluid and solid domain using mathematical homogenization. The presented model consists of heat equations inside the different domains, coupled through either perfect or imperfect thermal contact. The size and the period of the grains is of order $\varepsilon$, therefore forming a thin layer. The equation parameters inside the grains also depend on $\varepsilon$. We consider two distinct scenarios: Case (a), where the grains are disconnected, and Case (b), where the grains form a connected geometry but in a way such that the fluid and solid are still in contact. In both cases, we determine the effective differential equations for the limit $\varepsilon \to 0$ via the concept of two-scale convergence for thin layers. We also present and study a numerical algorithm to solve the homogenized problem.
△ Less
Submitted 21 December, 2023; v1 submitted 5 December, 2023;
originally announced December 2023.
-
Effective Heat Transfer Between a Porous Medium and a Fluid Layer: Homogenization and Simulation
Authors:
Michael Eden,
Tom Freudenberg
Abstract:
We are investigating the effective heat transfer in complex systems involving porous media and surrounding fluid layers in the context of mathematical homogenization. We differentiate between two fundamentally different cases: Case (a), where the solid part of the porous media is assumed to consist of disconnected inclusions and Case (b), where the solid matrix is assumed to be connected. For both…
▽ More
We are investigating the effective heat transfer in complex systems involving porous media and surrounding fluid layers in the context of mathematical homogenization. We differentiate between two fundamentally different cases: Case (a), where the solid part of the porous media is assumed to consist of disconnected inclusions and Case (b), where the solid matrix is assumed to be connected. For both scenarios, we consider a heat equation with convection where a small scale parameter epsilon characterizes the heterogeneity of the porous medium and conduct a limit process via two-scale convergence for the solutions of the epsilon-problems. In Case (a), we arrive at a one-temperature problem exhibiting a memory term and, in Case (b), at a two-phase mixture model. We compare and discuss these two limit models with several simulation studies both with and without convection.
△ Less
Submitted 30 April, 2024; v1 submitted 19 December, 2022;
originally announced December 2022.
-
Deep Learning Methods for Partial Differential Equations and Related Parameter Identification Problems
Authors:
Derick Nganyu Tanyu,
Jianfeng Ning,
Tom Freudenberg,
Nick Heilenkötter,
Andreas Rademacher,
Uwe Iben,
Peter Maass
Abstract:
Recent years have witnessed a growth in mathematics for deep learning--which seeks a deeper understanding of the concepts of deep learning with mathematics and explores how to make it more robust--and deep learning for mathematics, where deep learning algorithms are used to solve problems in mathematics. The latter has popularised the field of scientific machine learning where deep learning is app…
▽ More
Recent years have witnessed a growth in mathematics for deep learning--which seeks a deeper understanding of the concepts of deep learning with mathematics and explores how to make it more robust--and deep learning for mathematics, where deep learning algorithms are used to solve problems in mathematics. The latter has popularised the field of scientific machine learning where deep learning is applied to problems in scientific computing. Specifically, more and more neural network architectures have been developed to solve specific classes of partial differential equations (PDEs). Such methods exploit properties that are inherent to PDEs and thus solve the PDEs better than standard feed-forward neural networks, recurrent neural networks, or convolutional neural networks. This has had a great impact in the area of mathematical modeling where parametric PDEs are widely used to model most natural and physical processes arising in science and engineering. In this work, we review such methods as well as their extensions for parametric studies and for solving the related inverse problems. We equally proceed to show their relevance in some industrial applications.
△ Less
Submitted 16 May, 2023; v1 submitted 6 December, 2022;
originally announced December 2022.
