Course syllabus for Turbulence modeling

The course syllabus is not adopted.

Overview

  • Swedish nameTurbulensmodellering
  • CodeMTF271
  • Credits7.5 Credits
  • OwnerMPAME
  • Education cycleSecond-cycle
  • Main field of studyMechanical Engineering
  • DepartmentMECHANICS AND MARITIME SCIENCES
  • GradingTH - Pass with distinction (5), Pass with credit (4), Pass (3), Fail

Course round 1

  • Teaching language English
  • Application code 03117
  • Open for exchange studentsYes

Credit distribution

0120 Written and oral assignments, part A 1.5 c
Grading: UG		 1.5 c
0220 Written and oral assignments, part B 1.5 c
Grading: UG		 1.5 c
0320 Examination 4.5 c
Grading: TH		 4.5 c

In programmes

Examiner

Eligibility

General entry requirements for Master's level (second cycle)
Applicants enrolled in a programme at Chalmers where the course is included in the study programme are exempted from fulfilling the requirements above.

Specific entry requirements

English 6 (or by other approved means with the equivalent proficiency level)
Applicants enrolled in a programme at Chalmers where the course is included in the study programme are exempted from fulfilling the requirements above.

Course specific prerequisites

MTF073 - Computational fluid dynamics: The finite volume method (CFD) eller TME226Mechanics of fluid sor any corresponding course

Aim

The object of the course is to give the students a thorough knowledge and understanding of modern,
advanced turbulence models for unsteady fluid flow simulations.

Learning outcomes (after completion of the course the student should be able to)

  • describe different RANS turbulence models such as Reyonlds stess models, algebraic Reynolds stress models, k-eps, k-omega, V2F, k-omega SST.
  • understand and outline the difference between LES, RANS, URANS, DES and hybrid LES-RANS.
  • derive the exact transport turbulence equations using tensor notation.
  • describe the modeling assumptions, using tensor notation, in turbulence models.
  • identify and interpret the different terms in the turbulence models presented in the course.
  • understand the basics in machine learning.
  • describe the difference between resolved and modeled Reynolds stresses.
  • describe different approaches to handle the near-wall problem in LES.
  • describe the advantages of second-moment closures compared to eddy-viscosity models.
  • derive the V2F model.
  • reproduce the different spatial filtering approaches in LES.
  • derive the SGS models using tensor notation.
  • understand and describe the concept of modeled (for example SGS) dissipation between resolved and modeled scales.
  • describe the method for how to prescribe unsteady, fluctuating inlet boundary conditions.
  • be able to carry out an simulation with a commercial CFD code.

Content

The development of computers and Computational Fluid Dynamics (CFD) has made the numerical simulation of complex fluid flow, combustion, aero-acoustics and heat transfer problems possible. Turbulent flow in three-dimensional, complex geometries -- unsteady or steady -- can be dealt with. Presently CFD methods can replace, or complement, many experimental methods; we can use a numerical wind tunnel instead of an experimental one.

Today, most CFD simulations are carried out with traditional RANS (Reynolds-Averaged Navier-Stokes). In RANS, we split the flow variables into one time-averaged (mean) part and one turbulent part. The latter is modelled with a turbulence model such as k-eps or Reynolds Stress Model. For many flows it is not appropriate to use RANS, since the turbulent part can be very large and of the same order as the mean. Examples are unsteady flow in general, wake flows or flows with large separation. For this type of flows, it is more appropriate to use Large Eddy Simulation (LES). In order to extend LES to high Reynolds number flows new methods have been developed. These are called DES (Detached Eddy Simulation), URANS (Unsteady RANS) or Hybrid LES-RANS. They are all unsteady methods and they are a mixture of LES and RANS. In aero-acoustics the noise is generated by turbulence. The best way to accurately predict large-scale turbulence is to carry out an unsteady simulation of the flow field (i.e. LES, DES, hybrid LES-RANS or URANS). After that the noise is predicted separately in CAA (Computational Aero-Acoustics).

In LES, DES, URANS and Hybrid LES-RANS the large-scale part of the turbulence is solved for by the discretized equations whereas the small-scale turbulence is modeled. The definition of ''large-scale'' varies in the different methods. Furthermore, the limit between ''large-scale'' and ''small-scale' is often not well defined. Since turbulence is three-dimensional and unsteady, it means that in all the methods the simulations must always be carried out as three-dimensional, unsteady simulations.

We will address questions like:
  • How should I make my mesh?
  • Why should I in LES use a non-dissipative discretization scheme?
  • Is it necessary to use central differencing in DES and URANS?
  • What is the different between LES and unsteady RANS?
  • What turbulence models can I use in DES and unsteady RANS?
  • To enhance numerical stability, can a turbulence model with high dissipation be used?
  • How do I prescribe inlet boundary conditions?
  • Inlet boundary conditions: can I use steady inlet boundary conditions? Which is best, synthesized turbulence or a pre-cursor DNS?
In the first project, we will learn how to interpret results from an unsteady simulation. We will also use machine learning to improve a turbulence model.

When doing LES-URANS/DES, you have to ask yourself similar questions as when doing measurements:
  • When is the flow fully developed so that I can start time-averaging?
  • For how long time do I need to time-average?
  • Is it enough if I get accurate mean flow or do I also need accurate resolved turbulent stresses?
  • How do I estimate the quality of my LES or hybrid LES-RANS? Spectra? 2-point correlations? SGS dissipation?
The most important drawback/bottleneck of LES is the requirement to use very fine grid near walls. The grid must be fine in all directions, not only the wall-normal direction. Much of the research on LES is today focused in getting around this bottleneck. One approach is hybrid LES-RANS. In this method RANS is used near walls and LES is used in the remaining part of the domain.



Organisation


11 prerecorded lectures are uploaded to Canvas.

There will be five discussion seminars  during  lectures on Campus.
  • The students will register in groups (8-10 in each).
  •  Each discussion seminar will last 45 minutes.
 Two assignments (Assignment 1 and Assignment 2a and  2b) should be carried out by the students. The assignments will supervised on Campus. Students can use
  • Python (recommended), (recommended),
Both Octave and Python are open-source software. Many large Swedish industries prefer engineers to use Python instead of Matlab due to Matlab's high license fees.
For Assignment 2b, only Python scripts are available.


Two projects should be carried out by the students.

1. In the first part of Assignment 1, the students will be given data from a numerical simulation (LES or DNS). The data will be two-dimensional, time-averaged velocity (recirculating flow) and pressure fields, the Reynolds stresses and the (SGS) dissipation. The data will be analyzed.

We will start  analyzing the transport equations of the turbulent Reynolds stresses, u_iu_j. We identify regions of large production terms, which should correspond to regions of large Reynolds stresses. Reynolds stresses will be computed using the
eddy-viscosity assumption, and these will be compared to their exact counter-parts.


In the second part of Assignment 1, the students will use Machine Learning and try to improve turbulence models. The first attempt could be to improve the standard k-eps model my optimizing the C_mu coefficient. Influence parameters may be velocity gradients (coordinate-invariant) and/or the  turbulent time-scale, k/ε, both functions of x and y. The output parameter will be C_mu = C_mu (x,y). For more details, see  Course Plan

2. In Assignment 2, the students will be given flow fields of channel flow at high Reynolds number and the flow
over a hump. Both flows have been obtained using PANS/DES/IDDES.These flows will be analyzed using different modeling assumptions such as DES, DDES, PANS and SAS. For more detail, see Assignment 2a and 2b in the eBook  Course home page

Literature

The eBook can be downloaded from the course home page

Examination including compulsory elements

 Grades. failed, passed; grade 4; grade 5
  •  Part 1: Two assignments including written presentations. This part is mandatory.
  •  Part 2: Discussion seminars. This part is not mandatory.
  •  Part 3: Oral exam is mandatory.
  • Oral exam based on the questions in the Discussion seminars and the Assignments.The teachers will also ask follow-up questions.There we try to test if the student has understood the topic or if he/she has  memorized it. A good understanding gives grade 4 or 5. Two students at the time. .
  • To get grade 'passed', you must have passed the Oral exam and get passed on the two assignment reports.

The course examiner may assess individual students in other ways than what is stated above if there are special reasons for doing so, for example if a student has a decision from Chalmers about disability study support.