Appendix G. Automated thermal model correlation. Martin Trinoga (Airbus Safran Launchers, Germany)

Transcription

1 77 Appendix G Automated thermal model correlation Martin Trinoga (Airbus Safran Launchers, Germany)

2 78 Automated thermal model correlation Abstract An essential part in the development of a spacecraft is the establishment of a thermal model for the thermal design and the temperature predictions during the operation phase. In order to improve the accuracy of the temperature predictions, all thermal spacecraft models need to be correlated with measurements from thermal tests or if available thermal flight data. Since 013 a new tool for an automated model correlation is under development with the name TAUMEL. With this newly developed tool written in MATLAB(R) programming language it will be possible to correlate thermal models from ESATAN-TMS automatically. For the validation several different thermal models from real flight hardware such as various test models were used. A first insight was already given in the 8th European Space Thermal Analysis Workshop in October 014. Within the last two years a significant progress was mastered. Especially the functionality and user-friendliness was improved in the last period and lead to a promising tool for an effective automated thermal model correlation. On this conference the latest status, highlights and results which were treated during the past development will be presented.

3 Automated thermal model correlation 79 Automated thermal model correlation Martin Trinoga Airbus Safran Launchers GmbH Bremen / Germany th european space thermal analysis ESTEC Content Development, implementation and testing of an automated thermal model correlation tool TAUMEL* Why we did it Thermal model and correlation to test data How we solved it The correlation method for TAUMEL How we did it The implementation of the method in TAUMEL How good we did it The testing of TAUMEL What we still have to do Next steps * TAUMEL = Tool for AUtomated Model Correlation using Equation Linarization

4 80 Automated thermal model correlation 1 Why we did it Thermal model and correlation to test data 3 1) Why we did it TMM Goodness of fit Test T model,1 T model,n T test,1 T test,n Iterative Process Correlation: Modification of a thermal model (TMM) to fit the node temperatures and the measurement temperatures 4

5 Automated thermal model correlation 81 How we solved it The correlation method for TAUMEL 5 ) How we solved it (our solution) Parameterized model 4 GL 34 3 DOMAIN Initial TMM GL 13 GR 3 1 GF 1 QS, QA, QE, QR 6

6 8 Automated thermal model correlation ) How we solved it (our solution) Parameterized model 4 PL 34 GL 34 3 DOMAIN Modified TMM PL 13 GL 13 PR 3 GR 3 1 PF 1 GF 1 QS, QA, QE, QR Variation of parameters Variation of conductors 7 ) How we solved it (our solution) Linearized parameterized solver 1 C i T i = GL ij (T j T i ) + σ GR ik (T k ) 4 (T i ) 4 i j i k Read the governing equation solved by the TMM Selection of a time step t 0 + Q B,i i C i T i r = PL ij GL ij (T j T i ) i j Introduction of Parameters by the user + σ PR ik GR ik (T k ) 4 (T i ) 4 i k + Q B,i i 3 0 = PL ij GL ij (T j T i ) + σ PR ik GR ik T k T i i j i k + Q B,i + Q T + Q L i Linearization of the equation and opimization of Parameters 8

7 Automated thermal model correlation 83 3 How we did it The implementation of the method in TAUMEL 9 3) How we did it (Implementation) Case Case 3 TAUMEL 1 n 1 T T T M min. Case 1 T T T M min. Parameter allocation & grouping 4 PL34GL34 3 DOMAIN Automated TMM modification PL13GL13 PR3GR3 1 PF1GF1 QS, QA, QE, QR TAUMEL 10

8 84 Automated thermal model correlation 4 How good we did it The testing of TAUMEL 11 4) How good we did it (Testing) Test model 1N-Thruster Measurements coverage of 66% of model nodes Already existing conventional correlation for comparison Small model: o 45 diffusion nodes o 34 conductors 1

9 Automated thermal model correlation 85 4) How good we did it (Testing) 13 4) How good we did it (Testing) 14

10 86 Automated thermal model correlation 4) How good we did it (Testing) 15 4) How good we did it (Testing) Correlation Strategy Run Description of Strategy MAD [K] 0 13,48 1 All linear conductors GL ( ) 7,4 All GLs ,3 3 All GLs ,53 4 All GLs +- 0.,98 5 Nodes 40 and 70: GRs added to variation,60 6 All GLs and selected GRs +- 0.,45 7 All GLs and selected GRs +- 0.,30 8 Node 0: GRs on node 0 added to variation,9 9 All GLs and selected GRs +- 0.,10 10 All GLs and selected GRs ,86 11 All GLs and selected GRs ,73 1 All GLs and selected GRs ,60 16

11 Parameter value [-] Mean Absolute Deviation [K] Automated thermal model correlation 87 4) How good we did it (Testing) Evolution of goodness of fit MAD evolution for repeated TAUMEL runs N of TAUMEL run 17 4) How good we did it (Testing),5 Evolution of parameters Evolution of parameter values for repeated TAUMEL runs 1,5 1 0, N of TAUMEL run Parameter Parameter 60 Parameter 7 Parameter 14 Parameter Parameter 31 Parameter 4 Parameter 66 18

12 88 Automated thermal model correlation 4) How good we did it (Testing) Benchmarking Model Number of Number of Cycles Iterations Number of Nodes Processing Time [s] 1-N-Thruster N-Thruster ESC-A A5ME US ~3000 ~1h Experience: ~10 repetitions of the correlation for optimized results Model Preparation* Process Total correlation time [min] [min] (assessed) 1-N-Thruster 60 8 ~1h 0-N-Thruster 90 0 ~h ESC-A ~3h A5ME ~15h Nota: 1-N-Thruster conventional correlation ~4 Weeks of iterations 19 5 What we still have to do Next steps 0

13 Automated thermal model correlation 89 5) Next steps Test of TAUMEL from non-involved people (only with user-manual). Enabling to correlate Systema/Thermica models with TAUMEL. Code documentation e.g. flow chart. Function for assessment of conductor correlation impact. Documentation of current verification, validation and results. External usage of TAUMEL ASL is interested in external usage! However, we want to be involved in the correlation process in order to see if new thermal models lead to new unknown problems. No verification according to commercial software foreseen. 1

14 90 Automated thermal model correlation 6 Backup Slides 3 6) Backup Slides Uncorrelated TMM Test/flight data New approach overview Correlation iterations at selected time Test/flight data Correlated TMM Test/flight data Goodness of fit Goodness of fit ESATAN TMM TAUMEL in MATLAB ESATAN TMM *.TMD file PL ij PR ij PF ij Simulated Annealing Best parameters 4

15 Automated thermal model correlation 91 6) Backup Slides TAUMEL GUI 5 6) Backup Slides Model Calculation times Without GR improvementterm Without GR improvement-term ( calculations) ESATAN-TMS (transient calculation) 1-N-Thruster (45 nodes) s s 38.3s VINCI (170 nodes) s s 7m, 9s ESC-A (1467 nodes) s s m, 36s A5ME (~3000 nodes) s s >4h Slide 19 6

16 9