Search

This paper explores the potential of Physics-Informed Neural Networks (PINNs) to serve as Reduced Order Models (ROMs) for simulating the flow field within stirred tank reactors (STRs). We solve the two-dimensional stationary Navier-Stokes equations within a geometrically intricate domain and explore methodologies that allow us to integrate additional physical insights into the model. These approaches include imposing the Dirichlet boundary conditions (BCs) strongly and employing domain decomposition (DD), with both overlapping and non-overlapping subdomains. We adapt the Extended Physics-Informed Neural Network (XPINN) approach to solve different sets of equations in distinct subdomains based on the diverse flow characteristics present in each region. Our exploration results in a hierarchy of models spanning various levels of complexity, where the best models exhibit l1 prediction errors of less than 1% for both pressure and velocity. To illustrate the reproducibility of our approach, we track the errors over repeated independent training runs of the best identified model and show its reliability. Subsequently, by incorporating the stirring rate as a parametric input, we develop a fast-to-evaluate model of the flow capable of interpolating across a wide range of Reynolds numbers. Although we exclusively restrict ourselves to STRs in this work, we conclude that the steps taken to obtain the presented model hierarchy can be transferred to other applications., Comment: 39 pages, 19 figures

Medical therapy for patients with severe heart disease often relies on left ventricular assist devices (LVADs). It is an ongoing topic of research how complications like stagnation and thrombosis can be prevented by using artificial pulsatility. We study a parametric pulse profile and its effect on the left ventricular washout using a computational flow model. Based on finite element simulations, we construct reduced-order models to reduce the computational costs for evaluating the parametrized problem. We observe that a pulsatile flow rate improves the washout compared to a constant flow rate. In particular, we find that short, intense pulses wash out the left ventricle best, whereas the time between subsequent pulses plays no significant role.

We present a novel application of neural networks to design improved mixing elements for single-screw extruders. Specifically, we propose to use neural networks in numerical shape optimization to parameterize geometries. Geometry parameterization is crucial in enabling efficient shape optimization as it allows for optimizing complex shapes using only a few design variables. Recent approaches often utilize CAD data in conjunction with spline-based methods where the spline's control points serve as design variables. Consequently, these approaches rely on the same design variables as specified by the human designer. While this choice is convenient, it either restricts the design to small modifications of given, initial design features - effectively prohibiting topological changes - or yields undesirably many design variables. In this work, we step away from CAD and spline-based approaches and construct an artificial, feature-dense yet low-dimensional optimization space using a generative neural network. Using the neural network for the geometry parameterization extends state-of-the-art methods in that the resulting design space is not restricted to user-prescribed modifications of certain basis shapes. Instead, within the same optimization space, we can interpolate between and explore seemingly unrelated designs. To show the performance of this new approach, we integrate the developed shape parameterization into our numerical design framework for dynamic mixing elements in plastics extrusion. Finally, we challenge the novel method in a competitive setting against current free-form deformation-based approaches and demonstrate the method's performance even at this early stage.

Profile extrusion is a continuous production process for manufacturing plastic profiles from molten polymer. Especially interesting is the design of the die, through which the melt is pressed to attain the desired shape. However, due to an inhomogeneous velocity distribution at the die exit or residual stresses inside the extrudate, the final shape of the manufactured part often deviates from the desired one. To avoid these deviations, the shape of the die can be computationally optimized, which has already been investigated in the literature using classical optimization approaches. A new approach in the field of shape optimization is the utilization of Reinforcement Learning (RL) as a learning-based optimization algorithm. RL is based on trial-and-error interactions of an agent with an environment. For each action, the agent is rewarded and informed about the subsequent state of the environment. While not necessarily superior to classical, e.g., gradient-based or evolutionary, optimization algorithms for one single problem, RL techniques are expected to perform especially well when similar optimization tasks are repeated since the agent learns a more general strategy for generating optimal shapes instead of concentrating on just one single problem. In this work, we investigate this approach by applying it to two 2D test cases. The flow-channel geometry can be modified by the RL agent using so-called Free-Form Deformation, a method where the computational mesh is embedded into a transformation spline, which is then manipulated based on the control-point positions. In particular, we investigate the impact of utilizing different agents on the training progress and the potential of wall time saving by utilizing multiple environments during training., Comment: 35 pages, 21 figures, submitted to Advances in Computational Science and Engineering (ACSE)

We introduce a numerical workflow to model and simulate transient close-contact melting processes based on the space-time finite element method. That is, we aim at computing the velocity at which a forced heat source melts through a phase-change material. Existing approaches found in the literature consider a thermo-mechanical equilibrium in the contact melt film, which results in a constant melting velocity of the heat source. This classical approach, however, cannot account for transient effects in which the melting velocity adjusts itself to equilibrium conditions. With our contribution, we derive a model for the transient melting process of a planar heat source. We iteratively cycle between solving for the heat equation in the solid material and updating the melting velocity. The latter is computed based on the heat flux in the vicinity of the heat source. The motion of the heated body is simulated via the moving mesh strategy referred to as the Virtual Region Shear-Slip Mesh Update Method, which avoids remeshing and is particularly efficient in representing unidirectional movement. We show numerical examples to validate our methodology and present two application scenarios, a 2D planar thermal melting probe and a 2D hot-wire cutting machine.

A key step during industrial design is the passing of design information from computer aided design (CAD) to analysis tools (CAE) and vice versa. Here, one is faced with a severe incompatibility in geometry representation: While CAD is usually based on surface representations, analysis mostly relies on volumetric representations. The forward pass, i.e., converting CAD data to computational meshes, is well understood and established. However, the same does not hold for the inverse direction, i.e., CAD reconstruction of deformed geometries resulting, e.g., from shape optimization. The few reconstruction methods reported mainly rely on spline fitting, in particular methods that rely on creating new splines simililar to shape reconstruction from 3D imaging. In contrast, this paper studies a novel approach that reuses the CAD data given in the initial design. We show that this concept leads to precise shape reconstructions while also preserving the initial notion of features defined during design. Furthermore, reusing the initial CAD representation reduces the shape reconstruction problem to a shape modification problem. We study this unique feature and show that it enables the reconstruction of CAD data from computational meshes by composing each spline in the initial CAD data with a single, global deformation spline. While post-processing is needed for use in current CAD software, this novel approach not only allows creating watertight models, but also enables reconstructing complete CAD models even from defeatured computational meshes., Comment: 15 pages, 12 figures

One approach with rising popularity in analyzing time-dependent problems in science and engineering is the so-called space-time finite-element method that utilized finiteelements in both space and time. A common ansatz in this context is to divide the mesh in temporal direction into so-called space-time slabs, which are subsequently weakly connected in time with a Discontinuous Galerkin approach. The corresponding jumpterm, which is responsible for imposing the weak continuity across space-time slabs can be challenging to compute, in particular in the context of deforming domains. Ensuring a conforming discretization of the space-time slab at the top and bottom in time direction simplifies the handling of this term immensely. Otherwise, a computationally expensive and error prone projection of the solution from one time-level to another is necessary. However, when it comes to simulations with deformable domains, e.g. for free-surface flows, ensuring conforming meshes is quite laborious. A possible solution to this challenge is to extrude a spatial mesh in time at each time-step resulting in the so-called time-discontinuous prismatic space-time (D-PST) method. However, this procedure is restricted to finite-elements of 1st order in time. We present a novel algorithmic approach for arbitrarily discretized meshes by flipping the mesh in time-direction for each time-step. This ansatz allows for a simple evaluation of the jump-term as the mesh is always conforming. The cost of flipping the mesh around its symmetry plane in time scales with the number of nodes, which makes it computationally cheaper than an additional update of the mesh to enforce conformity or the evaluation of a projection. We validate the approach on various physical problems with and without deforming domains.

This paper presents a novel shape-optimization technique for the design of mixing elements in single-screw extruders. Extruders enable the continuous production of constant-cross-section profiles. Equipped with one or several screw-shaped rotors, extruders transport polymer particles towards the outlet. Due to shear heating, melting is induced and a melt stream is created, which can be further processed. While many variants of multi-screw extruders exist, a significant share of all extrusion machines is made up of single-screw extruders due to their comparatively low operating costs and complexity. While the reduced complexity yields economic benefits, single-screw extruders' mixing capabilities, i.e., their ability to produce a melt with a homogeneous material and temperature distribution, suffer compared to multi-screw extruders. To compensate for this shortcoming, so-called mixing elements are added to the screw to enhance mixing by recurring flow reorientations. In view of the largely unintuitive flow characteristics of polymer melts, we present an optimization framework that allows designing these mixing elements numerically based on finite-element simulations of the melt flow. To reduce the computational demand required by shape optimization of a complete mixing section, we only focus on the shape optimization of a single mixing element. This paper presents advances in three aspects of numerical design: (1) A combination of free-form deformation and surface splines is presented, allowing to parameterize the mixing element's shape by very few variables. (2) The combination of this concept with a linear-elasticity-based mesh update method to deform the computational domain without the need for remeshing is demonstrated. (3) A simple yet robust and sensitive objective formulation to assess distributive mixing in laminar flows based on a measure for the interfacial area is proposed., Comment: 15 pages, 11 figures, submitted to Finite Elements in Analysis and Design

We present a strategy for the numerical solution of convection-coupled phase-transition problems, with focus on solidification and melting. We solve for the temperature and flow fields over time. The position of the phase-change interface is tracked with a level-set method, which requires knowledge of the heat-flux discontinuity at the interface. In order to compute the heat-flux jump, we build upon the ghost-cell approach and extend it to the space-time finite element method. This technique does not require a local enrichment of the basis functions, such as methods like extended finite elements, and it can be easily implemented in already existing finite element codes. Verification cases for the 1D Stefan problem and the lid-driven cavity melting problem are provided. Furthermore, we show a more elaborate 2D case in view of complex applications.

In this work, we present a novel method for the mesh update in flow problems with moving boundaries, the phantom domain deformation mesh update method (PD-DMUM). The PD-DMUM is designed to avoid remeshing; even in the event of large, unidirectional displacements of boundaries. The method combines the concept of two mesh adaptation approaches: (1) The virtual ring shear-slip mesh updatemethod (VR-SSMUM); and (2) the elastic mesh update method (EMUM). As in the VR-SSMUM, the PD-DMUMextends the fluid domain by a phantom domain; the PD-DMUM can thus locally adapt the element density. Combined with the EMUM, the PD-DMUMallows the consideration of arbitrary boundary movements. In this work, we apply the PD-DMUM in two test cases. Within the first test case, we validate the PD-DMUM in a 2D Poiseuille flow on a moving background mesh. Subsequently the fluid-structure interaction (FSI) problem in the second test case serves as a proof of concept. More, we stress the advantages of the novel method with regard to conventional mesh update approaches.

In recent years, numerical methods in industrial applications have evolved from a pure predictive tool towards a means for optimization and control. Since standard numerical analysis methods have become prohibitively costly in such multi-query settings, a variety of reduced order modeling (ROM) approaches have been advanced towards complex applications. In this context, the driving application for this work is twin-screw extruders (TSEs): manufacturing devices with an important economic role in plastics processing. Modeling the flow through a TSE requires non-linear material models and coupling with the heat equation alongside intricate mesh deformations, which is a comparatively complex scenario. We investigate how a non-intrusive, data-driven ROM can be constructed for this application. We focus on the well-established proper orthogonal decomposition (POD) with regression albeit we introduce two adaptations: standardizing both the data and the error measures as well as -- inspired by our space-time simulations -- treating time as a discrete coordinate rather than a continuous parameter. We show that these steps make the POD-regression framework more interpretable, computationally efficient, and problem-independent. We proceed to compare the performance of three different regression models: Radial basis function (RBF) regression, Gaussian process regression (GPR), and artificial neural networks (ANNs). We find that GPR offers several advantages over an ANN, constituting a viable and computationally inexpensive non-intrusive ROM. Additionally, the framework is open-sourced to serve as a starting point for other practitioners and facilitate the use of ROM in general engineering workflows.

For most finite element simulations, boundary-conforming meshes have significant advantages in terms of accuracy or efficiency. This is particularly true for complex domains. However, with increased complexity of the domain, generating a boundary-conforming mesh becomes more difficult and time consuming. One might therefore decide to resort to an approach where individual boundary-conforming meshes are pieced together in a modular fashion to form a larger domain. This paper presents a stabilized finite element formulation for fluid and temperature equations on sliding meshes. It couples the solution fields of multiple subdomains whose boundaries slide along each other on common interfaces. Thus, the method allows to use highly tuned boundary-conforming meshes for each subdomain that are only coupled at the overlapping boundary interfaces. In contrast to standard overlapping or fictitious domain methods the coupling is broken down to few interfaces with reduced geometric dimension. The formulation consists of the following key ingredients: the coupling of the solution fields on the overlapping surfaces is imposed weakly using a stabilized version of Nitsche's method. It ensures mass and energy conservation at the common interfaces. Additionally, we allow to impose weak Dirichlet boundary conditions at the non-overlapping parts of the interfaces. We present a detailed numerical study for the resulting stabilized formulation. It shows optimal convergence behavior for both Newtonian and generalized Newtonian material models. Simulations of flow of plastic melt inside single-screw as well as twin-screw extruders demonstrate the applicability of the method to complex and relevant industrial applications.

This paper provides the description of a novel, multi-purpose spline library. In accordance with the increasingly diverse modes of usage of splines, it is multi-purpose in the sense that it supports geometry representation, finite element analysis, and optimization. The library features reading and writing for various file formats and a wide range of spline manipulation algorithms. Further, a new efficient and objective-oriented algorithm for B-spline basis function evaluation is included. All features are available by a spline-type independent interface. The library is written in modern C++ with CMake as build system. This enables it for usage in typical scientific applications. It is provided as open-source library., Comment: 16 pages, 4 figures, submitted to Advances in Engineering Software

In recent years, new methods have been developed to synthesize complex porous and micro-structured geometry in a variety of ways. In this work, we take these approaches one step further and present these methods as an efficacious design paradigm. Specifically, complex micro-structure geometry can be synthesized while optimizing certain properties such as maximal heat exchange in heat exchangers, or minimal weight under stress specifications. By being able to adjust the geometry, the topology and/or the material properties of individual tiles in the micro-structure, possibly in a gradual way, a porous object can be synthesized that is optimal with respect to the design specifications. As part of this work, we exemplify this paradigm on a variety of diverse applications., Comment: 13 pages, 13 figures, Submitted to SPM 2019, Published in Computer-Aided Design, Volume 115, October 2019, Pages 23-33

This paper presents a novel spline-based meshing technique that allows for usage of boundary-conforming meshes for unsteady flow and temperature simulations in co-rotating twin-screw extruders. Spline-based descriptions of arbitrary screw geometries are generated using Elliptic Grid Generation. They are evaluated in a number of discrete points to yield a coarse classical mesh. The use of a special control mapping allows to fine-tune properties of the coarse mesh like orthogonality at the boundaries. The coarse mesh is used as a 'scaffolding' to generate a boundary-conforming mesh out of a fine background mesh at run-time. Storing only a coarse mesh makes the method cheap in terms of memory storage. Additionally, the adaptation at run-time is extremely cheap compared to computing the flow solution. Furthermore, this method circumvents the need for expensive re-meshing and projections of solutions making it efficient and accurate. It is incorporated into a space-time finite element framework. We present time-dependent test cases of non-Newtonian fluids in 2D and 3D for complex screw designs. They demonstrate the potential of the method also for arbitrarily complex industrial applications.

This paper describes an inverse shape design method for thermoelastic bodies. With a known equilibrium shape as input, the focus of this paper is the determination of the corresponding initial shape of a body undergoing thermal expansion or contraction, as well as nonlinear elastic deformations. A distinguishing feature of the described method lies in its capability to approximately prescribe an initial heterogeneous temperature distribution as well as an initial stress field even though the initial shape is unknown. At the core of the method, there is a system of nonlinear partial differential equations. They are discretized and solved with the finite element method or isogeometric analysis. In order to better integrate the method with application-oriented simulations, an iterative procedure is described that allows fine-tuning of the results. The method was motivated by an inverse cavity design problem in injection molding applications. Its use in this field is specifically highlighted, but the general description is kept independent of the application to simplify its adaptation to a wider range of use cases., Comment: 22 pages, 13 figures

This article considers the NURBS-Enhanced Finite Element Method (NEFEM) applied to the compressible Navier-Stokes equations. NEFEM, in contrast to conventional finite element formulations, utilizes a NURBS-based computational domain representation. Such representations are typically available from Computer-Aided-Design tools. Within the NEFEM, the NURBS boundary definition is utilized only for elements that are touching the domain boundaries. The remaining interior of the domain is discretized using standard finite elements. Contrary to isogeometric analysis, no volume splines are necessary. The key technical features of NEFEM will be discussed in detail, followed by a set of numerical examples that are used to compare NEFEM against conventional finite element methods with the focus on compressible flow., Comment: This is a pre-print of an article submitted for publication in Lecture Notes in Computational Science and Engineering

In this paper, we present the numerical solution of two-phase flow problems of engineering significance with a space-time finite element method that allows for local temporal refinement. Our basis is the method presented in [3], which allows for arbitrary temporal refinement in preselected regions of the mesh. It has been extended to adaptive temporal refinement that is governed by a quantity that is part of the solution process, namely, the interface position in two-phase flow. Due to local effects such as surface tension, jumps in material properties, etc., the interface can, in general, be considered a region that requires high flexibility and high resolution, both in space and in time. The new method, which leads to tetrahedral (for 2D problems) and pentatope (for 3D problems) meshes, offers an efficient yet accurate approach to the underlying two-phase flow problems., Comment: 13 pages, 14 Figures, 1 Table

We present a boundary-conforming space-time finite element method to compute the flow inside co-rotating, self-wiping twin-screw extruders. The mesh update is carried out using the newly developed Snapping Reference Mesh Update Method (SRMUM). It allows to compute time-dependent flow solutions inside twin-screw extruders equipped with conveying screw elements without any need for re-meshing and projections of solutions - making it a very efficient method. We provide cases for Newtonian and non-Newtonian fluids in 2D and 3D, that show mesh convergence of the solution as well as agreement to experimental results. Furthermore, a complex, unsteady and temperature-dependent 3D test case with multiple screw elements illustrates the potential of the method also for industrial applications.

Engineering design via CAD software relies on Non-Uniform Rational B-Splines (NURBS) as a means for representing and communicating geometry. Therefore, in general, a NURBS description of a given design can be considered the exact description. The development of isogeometric methods has made the geometry available to analysis methods Hughes et al. (2005). Isogeometric analysis has been particularly successful in structural analysis; one reason being the wide-spread use of two-dimensional finite elements in this field. For fluid dynamics, where three-dimensional analysis is usually indispensable, isogeometric methods are more complicated, yet of course not impossible, to apply in a general fashion. This paper describes a method that enables the solution of fluid-structure interaction with a matching spline description of the interface. On the structural side, the spline is used in an isogeometric setting. On the fluid side, the same spline is used in the framework of a NURBS-enhanced finite element method (extension of Sevilla et al. (2011)). The coupling of the structural and the fluid solution is greatly facilitated by the common spline interface. The use of the identical spline representation for both sides permits a direct transfer of the necessary quantities, all the while still allowing an adjusted, individual refinement level for both sides.

A novel method - the Virtual Ring Shear-Slip Mesh Update Method (VR-SSMUM) - for the efficient and accurate modeling of moving boundary or interface problems in the context of the numerical analysis of fluid flow is presented. We focus on cases with periodic straight-line translation including object entry and exit. The periodic character of the motion is reflected in the method via a mapping of the physical domain onto a closed virtual ring. Therefore, we use an extended mesh, where unneeded portions are deactivated to control the computational overhead. We provide a validation case as well as examples for the applicability of the method to 2D and 3D models of packaging machines.

The simulation of certain flow problems requires a means for modeling a free fluid surface; examples being viscoelastic die swell or fluid sloshing in tanks. In a finite-element context, this type of problem can, among many other options, be dealt with using an interface-tracking approach with the Deforming-Spatial-Domain/Stabilized-Space-Time (DSD/SST) formulation. A difficult issue that is connected with this type of approach is the determination of a suitable coupling mechanism between the fluid velocity at the boundary and the displacement of the boundary mesh nodes. In order to avoid large mesh distortions, one goal is to keep the nodal movements as small as possible; but of course still compliant with the no-penetration boundary condition. Standard displacement techniques are full velocity, velocity in a specific coordinate direction, and velocity in normal direction. In this work, we investigate how the interface-tracking approach can be combined with isogeometric analysis for the spatial discretization. If NURBS basis functions of sufficient order are used for both the geometry and the solution, both a continuous normal vector as well as the velocity are available on the entire boundary. This circumstance allows the weak imposition of the no-penetration boundary condition. We compare this option with an alternative that relies on strong imposition at discrete points. Furthermore, we examine several coupling methods between the fluid equations, boundary conditions, and equations for the adjustment of interior control point positions., Comment: 20 pages, 16 figures

In an effort to increase the versatility of finite element codes, we explore the possibility of automatically creating the Jacobian matrix necessary for the gradient-based solution of nonlinear systems of equations. Particularly, we aim to assess the feasibility of employing the automatic differentiation tool TAPENADE for this purpose on a large Fortran codebase that is the result of many years of continuous development. As a starting point we will describe the special structure of finite element codes and the implications that this code design carries for an efficient calculation of the Jacobian matrix. We will also propose a first approach towards improving the efficiency of such a method. Finally, we will present a functioning method for the automatic implementation of the Jacobian calculation in a finite element software, but will also point out important shortcomings that will have to be addressed in the future., Comment: 17 pages, 9 figures

We explore a vexing benchmark problem for viscoelastic fluid flows with the discontinuous Petrov-Galerkin (DPG) finite element method of Demkowicz and Gopalakrishnan [1,2]. In our analysis, we develop an intrinsic a posteriori error indicator which we use for adaptive mesh generation. The DPG method is useful for the problem we consider because the method is inherently stable---requiring no stabilization of the linearized discretization in order to handle the advective terms in the model. Because stabilization is a pressing issue in these models, this happens to become a very useful property of the method which simplifies our analysis. This built-in stability at all length scales and the a posteriori error indicator additionally allows for the generation of parameter-specific meshes starting from a common coarse initial mesh. A DPG discretization always produces a symmetric positive definite stiffness matrix. This feature allows us to use the most efficient direct solvers for all of our computations. We use the Camellia finite element software package [3,4] for all of our analysis., Comment: 20 pages, 18 figures, 6 tables

Fluid flow applications can involve a number of coupled problems. One is the simulation of free-surface flows, which require the solution of a free-boundary problem. Within this problem, the governing equations of fluid flow are coupled with a domain deformation approach. This work reviews five of those approaches: interface tracking using a boundary-conforming mesh and, in the interface capturing context, the level-set method, the volume-of-fluid method, particle methods, as well as the phase-field method. The history of each method is presented in combination with the most recent developments in the field. Particularly, the topics of extended finite elements (XFEM) and NURBS-based methods, such as Isogeometric Analysis (IGA), are addressed. For illustration purposes, two applications have been chosen: two-phase flow involving drops or bubbles and sloshing tanks. The challenges of these applications, such as the geometrically correct representation of the free surface or the incorporation of surface tension forces, are discussed., Comment: 75 pages, 34 figures

Subject of this paper is the derivation of a new constitutive law in terms of the logarithm of the conformation tensor that can be used as a full substitute for the 2D governing equations of the Oldroyd-B, Giesekus and other models. One of the key features of these new equations is that - in contrast to the original log-conf equations given by Fattal and Kupferman (2004) - these constitutive equations combined with the Navier-Stokes equations constitute a self-contained, non-iterative system of partial differential equations. In addition to its potential as a fruitful source for understanding the mathematical subtleties of the models from a new perspective, this analytical description also allows us to fully utilize the Newton-Raphson algorithm in numerical simulations, which by design should lead to reduced computational effort. By means of the confined cylinder benchmark we will show that a finite element discretization of these new equations delivers results of comparable accuracy to known methods., Comment: 21 pages, 5 figures