Computational Fluid Dynamics

Selected Research Projects

 

ALBORZ: an efficient lattice Boltzmann solver

Responsible person: Dr.-Ing. Seyed Ali Hosseini, Dr. Farshad Gharibi

ALBORZ is a robust solver based on the lattice Boltzmann formulation, developed in the group since 2013. It is used successfully to investigate a variety of flows and configurations, like laminar and turbulent particulate flows, combustion, flows in porous media, heat transfer, crystallization, medical flows, particle beds, etc. The code is written in C++ and parallelized using MPI. Many examples of validation and simulations have been published in international journals and presented at large conferences.

 Alborz 1PRECCINSTA

 

Figure: (left) Transport of large spherical particles in a turbulent flow for a Reynolds number of 5600, coloured by flow velocity. (right) Temperature isosurface (1100 K) for a turbulent flame in the PRECCINSTA burner.

 

Simulations for the crystal growth using lattice Boltzmann method

Responsible person: M.Sc. Qianyan Tan

Simulating crystal growth is of great importance in a wide variety of application areas, going from aircraft wing de-icing to enantiomer separation through preferential crystal growth. Modeling of crystal growth is about the phase change between a solid phase and a liquid phase separated by an interface. Phase-field models are powerful tools to track the interface evolution of the crystals. The aim of this work is to validate and use a phase-field solver based on the Lattice Boltzmann (LB) method combined with species, energy and flow filed. Then, simulating the growth habits of different types of crystals.

Figure: The evolution of the snow crystal in the atmosphere.

Figure:  The growth process of the mandelic acid crystals in the growth cell.

 

CFD-O: CFD-Based Optimization

Development of a New Software Library (OPtimization Algorithms Library++)

Responsible person: Dr.-Ing. László Daróczy, apl. Prof. Dr. Gábor Janiga

The Optimization Algorithms Library (or OPAL++ for short) is the department’s newest optimization and parameterization software. OPAL++ is completely portable, object-oriented, has a two-step parallelization with MPI and can be controlled by its user-friendly scripting language. OPAL++ has thus far successfully been used in multiple projects and has already automatically performed over 200,000 CFD calculations. In particular, the software provides support for:

- multiple operation systems (Mac OS X, Linux, Windows);
- coupling with many software types (PTC Creo, OpenFoam, Gambit, ANSYS-Fluent, StarCCM+, CFX, etc.);
- assistance in automatic mesh generation;
- optimization on heterogeneous systems (for example, Windows + Linux coupled);
- SSH & SFTP protocol for communication, synchronization between nodes;
- Single and multi-objective function optimization methods (OMOPSO, SPEA2, NSGA II, Firefly, FastGPA, Differential Evolution, Omni-Optimizer, etc.);
- Non-intrusive Polynomial Chaos Expansion for any distribution (with orthogonalization);
- Design-Of-Experiment methods (Sobol, Hammersley, Near-Orthogonal Latin Hypercube, etc.);
- Various Surrogate Methods (Radial Basis Function, Ordinary Kriging, Least Squares Fitting, etc..).

Figure: above: Kriging after optimization, below: Pareto front during optimization

Figure: automatic mesh generation using Gambit

See also our book: http://www.springer.com/de/book/9783540721529

CFD - DEM

Responsible person: Dr.-Ing. Kristin Kerst

The DEM model (DEM: Discrete Element Method) is a promising method, for example, in describing interactions of solid crystals in continuous crystallizers. Coupled with Computational Fluid Dynamics (CFD), it provides an opportunity for a detailed description of the crystal particles’ fluidization behavior.  The simulation is carried out on open-source software, CFDEMcoupling (OpenFOAM - LIGGGHTS - coupling) and is computationally very intense (7 hours of computing time on 64 2.1GHz quad processors for one physical second and 200,000 particles). To accompany the simulations, experiments on the real-life crystallization process are available through Prof. Andreas Seidel-Morgenstern’s research group at the Max Planck Institute Magdeburg. The research is being carried out as part of the DFG Priority Program 1679.

NS 1

Figure: CFD-DEM simulation of fluidized asparagine monohydrate crystals in the crystallizer (a) - without activated ultrasonic crusher, (b) - with activated ultrasonic crusher

Direct Numerical Simulations (DNS)

Responsible persons: Dr.-Ing. Abouelmagd Abdelsamie, Dr.-Ing. Cheng Chi

DINO is the department’s newest DNS software. DINO uses low-mach formulation for the simulation of turbulent reactive flow and two-phase flow, with the following methods:

- Efficient parallelization with MPI
- Spatial discretization by sixth-order finite differences;
- Explicit (fourth-order), semi-implicit (third-order Runge-Kutta), or implicit (wheel-5) temporal integration;
- Different physical, thermal, kinetic parameters and transport properties are calculated using Cantera and / or Eglib
- Two-phase flow is calculated for resolved or unresolved particles thanks to an Euler-Lagrange method.
- Complex geometries are taken into account thanks to an Immersed Boundary Method (IBM).

Application examples: jet flames, spray combustion, densely laden suspensions, transition, medical flow systems...

Two possibilities are available for the evaluation process:
- On-the-fly for Post Processing analysis and Feature Extraction of combustion and flow properties of Direct Numerical Simulation;
- Classical, file-based analysis of large data sets (parallelized Matlab library AnaFlame).

Animation: above: completely resolved particles (blue sphere) in a turbulent flow (white area shows the turbulent enstrophy); below: isosurface of vorticity (Taylor-Green vortex)

Animation: Isosurface of the mixing ratios in the turbulent jet (H2 / Air)

Research project RETERO: Reduktion von Tierversuchen zum Schädigungsrisiko bei Turbinenpassagen durch Einsatz von Roboterfischen, Strömungssimulationen und Vorhersagemodellen" (Förderkennzeichen 031L0152A)

Responsible person: M.Sc. Dennis Powalla

The main goal of the RETERO research project is to minimize the use of live-fish in injury risk assessment in downstream turbine passages. Within this context, a new method shall be developed. This method will be based on experiments, complemented with partly-autonomous robotic sensor fish surrogates, as well as on numerical simulations, featuring a risk assessment with help of fish behaviour models and computational fluid dynamics (CFD).There are two major approach paths to this project at the laboratory.

The first is to minimize the injury and mortality rate by developing a prediction model. This model should forecast the movement of a migrating fish in a complex 3D hydrodynamic field and account for possible contacts between fish body and surrounding structures, such as moving turbine blades.  The current approach is to develop a multi-agent system algorithm based on results provided by CFD simulations, where agents and associated behaviour response are derived from live-fish observations.The second part is involved in the development of a robotic fish surrogate. Forces acting on a moving fish body are assessed to gain information about the necessary forces of a propulsion device of a fish robot. The movement of the fish is realized by a morphing mesh model. The grid vertices morph to accommodate the body movement according to a predefined function in accordance to real fish motion characteristics.

FishMotion

Hemodynamic Data Assimilation

Verantwortlich: M.Sc. Franziska Gaidzik

The flow state in intracranial arteries can be estimated using various methods. Phase-Contrast Magnetic Resonance Imaging (PC-MRI) measurements can provide 4D flow information (time-resolved 3D spatial flow velocities), but with limited temporal and spatial resolution. Computational Fluid Dynamics (CFD) simulations can capture even finer intracranial arteries based on the vessel geometry, but numerical computations are strongly influenced by boundary and initial conditions. Data assimilation can combine experimental and numerical sources of data (i.e. prior knowledge) in a statistically appropriate fashion. The potential of such a data assimilation approach is to provide detailed information on vascular flow based on measurement data and physical principles, and to reduce the uncertainty in related estimates. Within the context of this sub-project, computational methods allowing for enhancement of measured data ranging below the temporal and spatial experimental resolution limits will be developed.

OverviewDA

CFD Simulation of a Solid-Liquid Counter-Current Screw Extractor

Responsible person: M.Sc. Annemarie Lehr

More efficient methods of extracting artemisinin from the leaves of the Artemisia annua plant via a solid-liquid extraction process are of growing interest as artemisinin is increasingly needed in antimalarial drug. A continuously operated counter-current process developed at the MPI Magdeburg requires strict safety procedures due to the required organic solvent toluene, resulting in expensive and time-consuming experimental tests. In this project, the observed extraction process shall be represented by coupling Computational Fluid Dynamics (CFD) and classical compartment models. Consequently, the advantages of both methods are used and effort and costs are reduced. Of decisive importance is the exact calculation and presentation of the solid and liquid residence time, as these strongly influence the reaction kinetics in the physical process. With the developed model, the process shall subsequently be optimized with regard to the yield of artemisinin.

Verteilung_SolventBlätter


 

Figure: Phase distribution in a solid-liquid counter-current screw extraction process by using CFD simulations

Fluid-Structure-Interaction

Responsible person: M.Sc. Samuel Voß

The realistic simulation of aneurysms is only possible if, in addition to the fluidic side, the fluid mechanical side (properties of the vessel wall) is taken into account. This very complex fluid-structure interaction (FSI for short) can be realized by coupling the Navier-Stokes equations (fluid) and the conservation of momentum (structure). Different methods and software tools are used to this end:
- Investigation of the interaction intensity by comparing two-way coupling with one-way clutch, as well as implementation without coupling
- Complex geometry discretization by finite volume and finite elements
- Simulative execution a) completely within STAR-CCM+/CD-adapco or b) by coupling STAR-CCM+ (fluid) with Abaqus FEA/SIMULIA (structure)
- Realization of explicit and implicit coupling
NS2

Figure: Streamlines and wall tension in an aneurysm

Large-Eddy-Simulation

Responsible person: apl. Prof. Dr. Gábor Janiga

This group uses large eddy simulation mainly for the study of transition or for flow with low turbulence intensity, for example in a nozzle from the Food & Drug Administration (FDA), an idealized medical device. These calculations were computed using an LES model in ANSYS Fluent. The calculation took 470 hours using 32 cores.

Animation: Q-criterion of a large eddy simulation for the FDA nozzle

Film Formation and Decay

Responsible person: Dr.-Ing. Thomas Hagemeier

Development of numerical models for the simulation of film formation and decay in practical configurations.

NS6

Figure: calculated film thickness during spray impingement on a simplified A-pillar geometry

Development and Improvement of Specific Numerical Methods

Responsible person: Prof. Dr.-Ing. Dominique Thévenin

  • Reduction of computation times for CFD-containing chemical reactions using dynamic systems theory (attractive manifolds, ILDM; flamelet-based, FPI methods).
  • Acceleration of CFD calculations: Employing parallel computers, dynamic adaptive mesh, improved convergence (multi-mesh methods).
  • Development of specific methods for the accurate description of boundary and initial conditions.
  • Coupling CFD with a description of the population dynamics for process engineering applications

Animation: Example of dynamic mesh adaptaption during the calculation of a flame using the program UGC+. The convergence is accelerated considerably by employing multigrid methods.

Last Modification: 26.02.2024 - Contact Person: Webmaster