Thermal Model Correlation Using Genetic Algorithms


Frédéric Jouffroy
EADS Astrium
Frederic dot Jouffroy at astrium dot eads dot net

James Etchells
James dot Etchells at esa dot int


The correlation of Thermal Mathematical Models (TMMs) with test data is a particularly challenging activity because it happens just before the final flight predictions when the project is on the critical path. Current methods for correlation of TMMs typically rely on a heuristic approach which requires a great deal of manual work. This often yields results which are barely satisfactory, in terms of accuracy, but extremely costly, in terms of engineering time.

Although current methods for TMM correlation are predominantly manual, there are, in principle, a variety of different optimisation algorithms that could be used to automate or semi-automate the process. Among these stochastic techniques, Genetic Algorithms (GA) seem well suited to the problem in question and, furthermore, they have already been used by Astrium (and others) for post-test mechanical model correlation.

In 2005-2006 a joint CNES-Astrium R&D activity was initiated with the objective of verifying the feasibility of correlating thermal models to test results using GA. The conclusions of this study were that, although promising results were obtained, more work was required to obtain a well established and generic methodology suitable for use on space projects.

Using the 2005-2006 CNES activity as a starting point, two further activities were intiated in 2006-2007, the current activity (ESA/ESTEC contract No. 19840/06/NL/PA) sought to:

  1. continue to assess the applicability of genetic algorithms for TMM correlation
  2. evaluate the possibility of using alternative optimisation techniques (either stochastic or deterministic)

Additionally, a second CNES-Astrium R&D activity was started to invesigate further the localisation of errors in TMMs

In order to draw on more experience in the field of optimisation, a cooperation was established between Astrium and the Laboratoire d’Optimisation Globale (DSNA/SDER/LOG), located at the École Nationale de l’Aviation Civile (ENAC) in Toulouse.

Material Available for Download

The following material is available for download on this website:

Document Summary for ESA Study
File Name Document Title
thcor-asf-005-rpt.pdf Bibliography study of optimisation methods with focus on genetic algorithm techniques w.r.t. post-test thermal model correlation problem
thcor-asf-006-tn.pdf Test plan & test models for verification of post-test thermal model correlation by using a stochastic global optimisation method
thcor-asf-013-tn-1.pdf Verification of feasibility of post-test thermal model correlation by using genetic algorithms: analysis of test plan results
GeneticAlgorithms.pdf Presentation given at the 2007 European Workshop on Thermal and ECLS Software

Material produced from the CNES R&D studies, mentioned above, may also become available on this web-page in the future.

The downloads below require a valid username and password. To obtain a username and password, please register here with your name, occupation and organisation. You will receive your username and password within 1 to 7 days usually.

Feedback and comments

Any feedback, comments or suggestions for future work relating to this study are much appreciated. Correspondence can be sent to the contact addresses at the top of this page and we will try to reply to you swiftly.

Terms and Conditions

The subject material was produced under an internal project of the European Space Agency. Therefore it is freely available to organisations and individuals located in one of the Agency's member or cooperating states according to the criteria set out in Article II (3) of Annex V of the European Space Agency Convention [*].

The subject material shall not be redistributed or published in any form outside the recipient's organisation. By downloading the material from this website you consent to these Terms and Conditions.

[*] Available from


Warranty and Disclaimer of Warranty. Although every attempt has been made to ensure the correctness of these models, they are provided "AS IS" and WITHOUT WARRANTY, either express or implied, including, without limitation, the warranties of NON-INFRINGEMENT, MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY OF THIS SOFTWARE IS WITH THE USER.  

Limitation of Liability. Under no circumstances and under no legal theory, whether in tort (including negligence), contract, or otherwise, shall the provider be liable to any person for any direct, indirect, special, incidental, or consequential damages of any character arising as a result of the use of these models including, without limitation, damages for loss of goodwill, work stoppage, computer failure or malfunction, or any and all other commercial damages or losses.