38th Advanced Computational Fluid Dynamics. Adjoint methods and their application in Computational Fluid Dynamics
Monday 14 September 2015 - Thursday 17 September 2015VKI secretariat, This email address is being protected from spambots. You need JavaScript enabled to view it.; Phone: +32 2 359 96 04
Introduction
Adjoint based sensitivity computations can be applied in many fields of modern CFD such as: error estimation for adapted grid generation (EE), multidisciplinary optimization (MDO) and uncertainty quantification (UQ). The latter application in UQ is relatively new and few developments exist. The purpose of this lecture series is to bring together the researchers from these different communities, and to stimulate the common interest and exchange of knowhow in using the adjoint formulation. Many times these methods are also coupled with state-of-the-art high order discretization methods such as Discontinuous Galerkin Finite Elements (DG), Finite Volumes (FV), Streamline Diffusion (SD) or Residual Distribution (RD). Moreover, there is also a strong potential in a combination of the different applications such as adaptive grids based on goal oriented error estimation combined with shape optimization, or uncertainty quantification combined with optimization, whereby the adjoint problem can be reused with a different forcing function in the right hand side.
The theory and algorithms used in the field of adjoints, error estimation, optimization and uncertainty quantification will be discussed in full detail. Emphasis is also put on industrial applications, in particular in the area of steady and unsteady aerodynamics. Achievements, weaknesses and perspectives for future research will be critically evaluated.
Who should attend ?
The course is oriented towards both junior and experienced engineers and researchers involved with CFD algorithm and code development. The lecture series assumes no preknowledge on adjoints nor on UQ, optimization and error estimation, but it is expected that the participants have a good background in CFD methods and numerical techniques for solving partial differential equations. This implies that the course will be a balanced mix between introductory and advanced lectures.
Time and Place
The course will take place at the von Karman Institute from Monday 14 September 2015 at 9:00 until Thursday 17 September 2015 at 17:00.
The Directors of the Lecture Series are Prof. H. Deconinck and Dr. T. Horvath of the von Karman Institute. The lecturers are top experts from both Europe and North America, namely:
- Prof. J. Alonso, Stanford U, USA (optimization)
- Prof. K. Fidkowski, U Michigan Ann Arbor, USA (error estimation)
- Prof. N. Gauger, Kaiserslautern U, Germany (optimization)
- Prof. G. May, RWTH Aachen U, Germany (error estimation)
- Dr. P. Congedo, INRIA Bordeaux Sud-Ouest (uncertainty quantification)
- Prof. D. Mavriplis, University of Wyoming, USA (uncertainty quantification and optimization)
Programme
Monday 14 September
- 08:45 Registration
- 09:00 Welcome Address
Prof. J. Muylaert, VKI Director - 09:15 Introduction to the adjoint method: derivations, uses, and properties
Prof. J. Alonso, Stanford U., USA - 10:30 Coffee break
- 11:00 Adjoint methods for sensitivity analysis and optimization
Prof. J. Alonso - 12:15 Lunch
- 14:00 Introduction to uncertainty quantification
Dr. P. Congedo, INRIA Bordeaux Sud-Ouest, France - 15:15 Coffee break
- 15:45 Applications of adjoint methods in aerodynamic shape optimization of supersonic aircraft
Prof. J. Alonso - 17:00 Welcome Drink
Tuesday 15 September
- 09:00 Advanced methods for uncertainty quantification and application of adjoints
Dr. P. Congedo - 10:15 Coffee break
- 10:45 Error Estimation and Mesh Adaptation Using Output Adjoints
Prof. K. Fidkowski, U of Michigan Ann Arbor, USA - 12:00 Lunch
- 14:00 Some perspectives on uncertainty-based derivative-free optimization
Dr. P. Congedo - 15:15 Coffee break
- 15:45 Output-based adaptive methods for unsteady flow problems
Prof. K. Fidkowski
Wednesday 16 September
- 09:00 Time dependent adjoint methods for single and multi-disciplinary problems
Prof. D. Mavriplis, U of Wyoming, USA - 10:15 Coffee break
- 10:45 Functional error estimation and control for time-dependent problems
Prof. D. Mavriplis - 12:00 Lunch
- 14:00 Hybridized discontinuous Galerkin methods: formulation and discrete adjoint
Prof. G. May, RWTH Aachen U, Germany - 15:15 Coffee break
- 15:45 Output-based error estimation and mesh adaptation for HDG methods
Prof. G. May
Thursday 17 September
- 09:00 Efficient AD-based Approaches for Optimization and Control in Aeronautics. Part I
Prof. N. Gauger, U of Kaiserslautern, Germany - 10:15 Coffee break
- 10:45 Adjoint-Based Error Estimation for Petrov-Galerkin Methods
Dr. S. D'Angelo, VKI, Belgium - 12:00 Lunch
- 14:00 Efficient AD-based Approaches for Optimization and Control in Aeronautics. Part II
Prof. N. Gauger - 15:15 Coffee break
- 15:45 Adjoint Methods for Uncertainty Quantification
Prof. D. Mavriplis
Fee & Registration
Extended Early bird (until 14 August 2015 included)
Normal fee (after 14 August 2015)
Type 1: Permanent residents of NATO countries funding VKI: Belgium, Czech Republic, France, Germany, Greece, Hungary, Iceland, Italy, Luxemburg, Norway, Portugal, Romania and Turkey
Type 2: Permanent residents of NATO countries not funding VKI or NATO partner countries
Type 3: Permanent residents of non -NATO countries
For PhD candidates and undergraduate students: The request to be considered for a discount must accompany the application to attend the Lecture Series, and the applicant must provide a recommendation letter from his or her professor; if not, the request will not be taken into consideration. All possible alternative sources of funding should be investigated before aid is requested under this scheme, so that those most in need will benefit.
REDUCTION for the normal fee. Not applicable when early bird.
-50% for the 3rd & 4th participant from the same company. The normal fee price list is available here.
METHOD OF PAYMENT
Payment 2 weeks prior to the beginning of the course (name and course title clearly indicated) by bank transfer to our account IBAN BE57 2100 3153 3035 , SWIFT BIC GEBABEBB
Location : von Karman Institute for Fluid Dynamics, Sint-Genesius-Rode (near Brussels), Belgium