|
Computing
Laboratory
Introduction:
The
incorporation of engineering simulation software into engineering design is
presently hindered by slow simulation turn-around times, poor gradient accuracy
and long adjoint code development times. In a number of cases we believe that
present AD tools would address such issues and provide valuable input
to AD researchers in developing the next generation of AD tools and
associated theory. Crucial to the above is having the educated manpower to use
the AD tools appropriately and intervening directly into the differentiated
code where necessary to improve efficiency.
Project
Aims:
-
To provide
efficient gradient code using AD for a complicated structural engineering
design problem and hence allow for a novel coupling of evolutionary and
gradient based optimisation. As well as underpinning new engineering design
strategies this will be published as a case study in obtaining AD
gradient code for a complex engineering application.
-
To carefully
compare the quality of hand-coded vs. AD-produced adjoint code for aerodynamic
adjoints. To date all discrete adjoint code has been developed either by
hand-coding or AD. Here we will uniquely compare a previously produced
state-of-the-art hand-coded adjoint solver with an AD produced version in
terms of both ease of development and run-time requirements.
-
To apply the
lessons learned from the steady flow solver of (ii) to an adjoint associated
with an unsteady-like, space-marched flow solver. Here we will further
investigate issues of check-pointing i.e., how much of the evolved flow field
should be stored and how much recomputed for the adjoint associated with an
implicit flow solver.
-
To clearly
show how to utilise AD to provide Jacobian-vector products and
preconditioning matrices within a Newton solver for Turbulent Navier-Stokes
flows. Hence we will demonstrate how the improved accuracy of forward
mode AD (compared to finite-differencing) overcomes the associated
doubled expense to result in overall improved efficiency and robustness. This
will extend present promising results (Hovland
& McInnes Parallel Computing 27, 2001) to a solver of greater complexity and poorer
numerical conditioning (stiff system, stretched mesh)
-
To thoroughly
educate a researcher as a U.K. AD application specialist through the
above tasks.
-
Feedback lessons learned back into the AD
community to improve AD tools and theory applied to such problems as
(1-4).
Project
Meetings:
-
14th May
2004, RMCS Shrivenham
-
23rd June
2005, RMCS Shrivenham
-
6th March
2006, DCMT Shrivenham
Journal
Papers:
-
Elimination
AD Applied to Jacobian Assembly for an Implicit Compressible CFD Solver, Mohamed Tadjouddine,
Shaun A Forth & Ning Qin, International Journal for Numerical Methods in Fluids , Volume 47 ,
Issue 10-11 (January 2005) pp 1315 - 1321. Special Issue: 8th ICFD Conference on Numerical
Methods for Fluid Dynamics . Issue Edited by Mike J. Baines, Mike J.P. Cullen, Chris
Farmer, Mike B. Giles, M. Rabbitt. Presented at The ICFD Conference for Numerical Methods in Fluid
Dynamics,
Oxford, 29th March-1st April 2004
-
Books:
Conference
Papers:
-
-
Adjoint
Differentiation of a Structural Dynamics Solver, Mohamed Tadjouddine, Shaun Forth &
Andy Keane in Automatic Differentiation: Applications, Theory, and
Implementations, Bücker, M.; Corliss, G.; Hovland,
P.; Naumann, U.; Norris, B. (Eds.)Lecture Notes in Computational Science & Engineering, Volume 50, p309-319, Springer, 2006. ISBN
3-540-28403-6
-
-
Novel
Passive Vibration Isolators,
A. I. J. Forrester, A. J. Keane, accepted ISMA International Conference on Noise and Vibration
Engineering, Leuven, September 2006.
(AD2CompEng generated code used in further EPSRC
funded research).
Conference
& Workshop Presentations:
-
Differentiating a
Time-Dependent CFD Solver, Emmanuel Tadjouddine, Shaun Forth & Ning
Qin, 1st European AD Workshop, April 2005, Nice, France.
-
Introduction
to Automatic Differentiation, Shaun Forth, Shorth Course in Automatic Differentiation,
AD2004 4th International Conference on Automatic Differentiation,
19-23 July 2004, University of Chicago, Gleacher Centre, Chicago, USA.
-
Automatic
Differentiation for a Structural Optimization Solver, Shaun Forth, Mohamed Tadjouddine & Andy Keane,
EUCCO2004: European Conference on Computational Optimization,
Minisymposium on Automatic Differentiation and Large Scale Optimization, March
29-31 2004, Technical University of Dresden, Germany.
AMOR home
|