UNCERTAINTY QUANTIFICATION ON A REAL CASE WITH TELEMAC-2D

Transcription

1 UNCERTAINTY QUANTIFICATION ON A REAL CASE WITH TELEMAC-2D Cédric Goeury, Thomas David, Riadh Ata, Sébastien Boyaval, Yoann Audouin, Nicole Goutal, Anne-Laure Popelin, Mathieu Couplet, Michael Baudin, Renaud Barate SALOME S USER DAY 26 NOV NOVEMBRE 2015 CEA 26 NOV 2015 PAGE 1

2 SOMMAIRE 1. Context and objectives 2. Problem specification 3. Computing environment 4. Uncertainty quantification using SALOME platform 5. Conclusions and Outlooks 27 NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 2

3 1. CONTEXT AND OBJECTIVES CEA 26 NOV 2015 PAGE 3 27 NOVEMBRE 2015

4 EVOLUTION OF DETERMINISTIC SIMULATIONS TOWARDS UNCERTAINTY Sizing Reliability (dams flooding risk heating sink warning system) regulatory demand (CMS, PPI, discharge ) Production optimization: Forecasting constraint related to the safety and regulation Complex studies in hydro-environmental fields Flooding risk Extreme Flow Optimizing hydro-electric production: Optimization and Forecasting Hydrology ATC Berre Lagoon Anticipate the risk of the clogging on the nuclear power plants: Forecasting 27 NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 4

5 HYDRO-ENVIRONMENTAL MODELLING COMPLEX PROCESS AND UNCERTAIN DATA Quantification and reduction of uncertainty optimization of parameters Forecasting Approximated model: TELEMAC-MASCARET system Not well-known parameters Uncertain datas: initial condition, boundary conditions, weather forcing, etc Observations : incomplete view of the process but not approximated Uncertainty Quantification : Impact of uncertainty on the output of the model sensitivity analysis etc Data Assimilation: Compromise between the approximated model and measurements to better simulate and forecast 27 NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 5

6 2. PROBLEM SPECIFICATION CEA 26 NOV 2015 PAGE 6 27 NOVEMBRE 2015

7 HYDRAULIC STUDY CASE North 27 NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 7

8 METHODOLOGIED FRAMEWORK FOR TREATING UNCERTAINTIES IN INDUSTRY 27 NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 8

9 2-D FREE SURFACE FLOW MODELLING Hypothesis: Uncompressible flow Hydrostatic pressure Impermeability of the surface and the bottom h div( hu) 0 t uuu vu g Z t x y x vuvv g Z t x y v y Advection s s F F Gradient of hydrostatic pressure x y 1 div( hegrad( u)) h 1 div( he grad( v)) h Diffusion Turbulence Dispersion Source terms, friction Uncertain factors Variable of interest 2 2 u v 4/3 2 F 1 cos( ) h g K u Boundary conditions: Upstream : Imposed discharge (Q) Downstream : free surface elevation imposed 27 NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 9

10 3. COMPUTING ENVIRONMENT CEA 26 NOV 2015 PAGE NOVEMBRE 2015

11 SALOME s User Day 26 NOV 2015 PAGE 11 COMPUTING ENVIRONMENT SALOME-HYDRO platform 27 NOVEMBRE 2015

12 HOW DOES IT WORK? Based on the interoperability of hydraulic models and OpenTURNS in SALOME-HYDRO Interoperability: Capacity of the software to run and share informations with other different softwares coupling compatibility Yacs module TELEMAC-2D wrapper 1 27 NOVEMBRE 2015 SALOME- HYDRO wrapper 2 OpenTURNS Using TELEMAC-2D as a node of YACS consists in creating a SALOME component: that drives the execution of the solver, and exhibits a service interface that can manage the interoperability with the environment SALOME (input/output data, parametric variables, and so on.) SALOME s User Day 26 NOV 2015 PAGE 12

13 4. UNCERTAINTY QUANTIFICATION USING SALOME PLATFORM CEA 26 NOV 2015 PAGE NOVEMBRE 2015

14 SALOME INTERFACE OF THE HYDRAULIC STUDY 27 NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 14

15 SALOME INTERFACE OF THE UNCERTAIN VARIABLES 27 NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 15

16 OpenTURNS module DEFINITION OF PROBABILISTIC STUDY Step 1: Define the distributions for our probabilistic variables 27 NOVEMBRE 2015 Q distribution Ks distribution Step 2: Associate those distributions with the variables of the model Step 3: Define the criteria for our study: Min/Max Central Uncertainty Threshold exceedence Yacs module The probabilistic study and variables is associated with the hydraulic computation model 27 NOVEMBRE NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 16

17 SALOME s User Day 26 NOV 2015 PAGE 17 UNCERTAINTY PROPAGATION METHOD OF MONTE CARLO Monte Carlo Simulations Q Propagation Output Ks Propagation Hydraulic model TELEMAC-2D generation of the ensemble of input random variables 1 1 n n n Q,..., K,..., x Q,..., K 1 x 5 5 Law of response: statistical estimators 1 ) n 1 n n j 2 1 j ˆ h h( x ˆ h h( x ) ˆ h j Central Limit Theorem: ˆ h h n N(0,1) n y n 1 j1 Convergence speed 2 1 n h x i 27 NOVEMBRE 2015

18 UNCERTAINTY PROPAGATION USING MONTE CARLO METHOD Global analysis local analysis Convergence and confidence intervals at 95 % of the estimated mean Empirical PDF based on simulations SALOME s User Day 26 NOV 2015 PAGE 18

19 SENSITIVITY ANALYSIS USING MONTE CARLO METHOD Global analysis : First order indices of Sobol local analysis S Q S CF4 S CF4 The flowrate imput factor explains about 80% of the variance of the output variable Few interactions between the uncertain variables 27 NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 19

20 CONCLUSIONS AND OUTLOOKS Feasibility of Monte Carlo approach on two-dimensional hydraulic studies using High Performing Computation The treatment of uncertainty in hydraulic studies is possible through interoperability hydraulic codes and uncertainties software OpenTURNS The coupling between TELEMAC-2D and OpenTURNS is facilitated in the platform Salome (OpenTURNS exists outside, and in SALOME) and it allows to access the features of SALOME platform (distribution of simulation on cluster calculations,...) Demonstrate the gain of uncertainty quantification in hydraulic applications, particularly in the context of 10 years visit 27 NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 20

21 THANK YOU FOR YOUR ATTENTION PAGE 21 CEA 26 NOV 2015 Direction Département Service 27 NOVEMBRE 2015

22 HOW DOES IT WORK? Based on the interoperability of hydraulic models and OpenTURNS in SALOME Interoperability: Capacity of the software to run and share informations with other different softwares coupling compatibility Have building blocks for each software that can be easily assembled to model complex systems TELEMAC 2D API: Control the execution of TELEMAC-2D Initialization, lecture of the imput datas, execution of a time step of computation Retrieve/Modify the current computation datales données du calculs en cours Number of mesh points, flowrate of one point, name of the result file, number of the time step and so on SALOME s User Day 26 NOV 2015 PAGE 22

23 WHAT ALLOW THE SALOME HYDRO API? Coupling facility: Multidomains (1D-2D for example) Multiphysics (water quality, sedimentology, etc.) Uncertainty Quantification Data Assimilation Shape optimization And so on Yacs module TELEMAC-2D wrapper 1 wrapper 2 OpenTURNS SALOME 27 NOVEMBRE NOVEMBRE 2015 SALOME s User Day 26 NOV 2015 PAGE 23