User day #17 2024

This year the OpenTURNS Users'Day will be held on Friday 14 june 2023 9h30-17h00 at EDF Lab Saclay.

EDF Lab Saclay is located in Boulevard Gaspard Monge, Palaiseau. See here to locate the site: https://www.google.fr/maps/place/EDF+Lab+Paris-Saclay/@48.7159083,2.1946009,17z/data=!3m1!4b1!4m6!3m5!1s0x47e678beda89e9cd:0xd5dc25575693ff25!8m2!3d48.7159083!4d2.1971758!16s%2Fg%2F11cjpck0pz

The slides will be in english and we will speak english if necessary.

To register: https://evenium.events/2nkxcpc3/

Program:

8h30-9h30: Breakfast

• 9h30-10h: Introduction, Y. Deleris (Airbus Research Manager)
• 10h-11h: Statistical Modelling of Extreme Values, S. Girard (INRIA Alpes)
• 11h-12h:
  • New Features of OpenTURNS 1.22 and 1.23 and of Persalys: J. Schueller, G. Garcia (Phimeca).
  • Sampling under constraints: R. Lebrun (Airbus)
  • New Features on extreme values modelling: A. Dutfoy (EDF)
  • Focus on the calibration capabilities of OpenTURNS : M. Baudin (EDF)

12h-13h30: Lunch with Persalys demonstration (G. Garcia)

• 13h30 - 15h45: Studies OpenTURNS inside (1/2)
  • Sensitivity Analyses of a Multi-Physics Long-Term Clogging Model For Steam Generators: V. Chabridon (EDF R&D)
  • Use of OpenTURNS for Uncertainty Treatment in Hydro-Sediment Modelling: F. Ben SAID (EDF R&D)
  • Automotive Reliability Engineering with OpenTURNS : the Phimeca product for Renault : StaRe (STAtistical REliability): G. Garcia (Phimeca), N. Bachelier (Renault)
  • OtFMI: standard container for exchange of dynamic simulation models: S.Girard (Phimeca)
  • Constraint Design of Experiments & Adaptive Sampling: I. Curtius (Airbus)

15h45 - 16h15: Gourmet break

• 16h15 - 17h25: Studies OpenTURNS inside (2/2)
  • Surrogate Modeling of High-Fidelity Radar Cross Section Simulation Using OpenTURNS PCE: T. Gricourt (ONERA)
  • otkerneldesign: an OpenTURNS module for kernel-based design of experiments: E. Fekhari (EDF R&D)
  • Benchmark of integrated solutions for uncertainty quantification: B. Kerleguer (CEA)

Presentations will be available at https://github.com/openturns/presentation#openturns-presentations

Abstracts:

Sensitivity Analyses of a Multi-Physics Long-Term Clogging Model For Steam Generators: V. Chabridon (EDF R&D)

Long-term operation of nuclear steam generators can result in the occurrence of clogging, a deposition phenomenon that may increase the risk of mechanical and vibration loadings on tube bundles and internal structures as well as potentially affecting their response to hypothetical accidental transients. To manage and prevent this issue, a robust maintenance program that requires a fine understanding of the underlying physics is essential. This study focuses on the utilization of a clogging simulation code developed by EDF R&D. This numerical tool employs specific physical models to simulate the kinetics of clogging and generates time dependent clogging rate profiles for particular steam generators. However, certain parameters in this code are subject to uncertainties. To address these uncertainties, Monte Carlo simulations are conducted to assess the distribution of the clogging rate. Subsequently, polynomial chaos expansions are used in order to build a metamodel while time-dependent Sobol’ indices are computed to understand the impact of the random input parameters throughout the whole operating time. Comparisons are made with a previous published study and additional Hilbert-Schmidt independence criterion sensitivity indices are computed. Key input-output dependencies are exhibited in the different chemical conditionings and new behavior patterns in high-pH regimes are uncovered by the sensitivity analysis. These findings contribute to a better understanding of the clogging phenomenon while opening future lines of modeling research and helping in robustifying maintenance planning.

Use of OpenTURNS for Uncertainty Treatment in Hydro-Sediment Modelling, F. Ben SAID (EDf R&D)

Hydro-sediment modelling is impacted by uncertainties from diverse sources. A key approach to understanding these uncertainties is sensitivity analysis, notably through Sobol’ indices, which help identify influential factors among model variables. However, traditional methods of estimating Sobol indices, such as the Saltelli (Monte Carlo) approach, can be computationally expensive. In my thesis work, I employ the Polynomial Chaos method (PCE) as a cost-efficient alternative for model approximation, addressing the computational challenges of sensitivity analysis. Despite the advantages of PCE, it encounters difficulties in accurately modelling discontinuous dynamics. To address this challenge, I developed a method to identify points of discontinuity within the model and proposed an adaptation of the Polynomial Chaos Expansion (PCE) method to handle these scenarios effectively. This adaptation enables the computation of Sobol indices efficiently using the PCE method even in the presence of discontinuities. Throughout my study, I exclusively rely on the OpenTURNS library, which offers robust tools and functionalities for uncertainty quantification and sensitivity analysis.

Automotive Reliability Engineering with OpenTURNS : the Phimeca product for Renault : StaRe (STAtistical REliability), G. Garcia (Phimeca), N. Bachelier (Renault)

Following 2023 presentation, best practices are now deployed to make most realistic Stress – Strength computations regarding customer usage. From usage (maneuvers / km), using parametric or non-parametric or upper tail distribution and usage-mileage dependency fitting, Stress distribution (maneuvers @ given mileage or time in service) is more realistic. Following Stress – Strength computations and no failures validation plan dimensioning give much more right sized tests. In 2024, Phimeca delivered the Python module StaRe, based on OpenTURNS. Major OpenTURNS-based module components will be described.

OtFMI: standard container for exchange of dynamic simulation models, S.Girard (Phimeca)

The OtFMI (https://github.com/openturns/otfmi) module provides to OpenTURNS users a seamless interface for using models in the Functional Mockup Unit (FMU) exchange format. An FMU is a container specified by the Functional Mockup Interface (FMI, https://fmi-standard.org/) akin to the wrappers familiar to the OpenTURNS’ community. Initiated by the Modelica Association community, the FMI standard is actively developed by a dedicated team and supported for import and/or export by more than 200 tools (https://fmi-standard.org/tools/). It has become the de facto leading standard to exchange dynamic simulation models, beyond the sole Modelica based models. Recent developments by the OpenTURNS team (Phimeca and EDF) added features to OtFMI for exporting OpenTURNS functions as FMU, as well as direct inclusion of arbitrary Python function in Modelica models. We will give an overview of the FMI–Python ecosystem and illustrate the potential of OtFMI through typical use cases.

Constraint Design of Experiments & Adaptive Sampling, E. Curtius (Airbus)

How to reduce the overall number of calculations necessary to describe a high dimensional design space by using Constraint Design of Experiments & Adaptive Sampling.

Surrogate Modeling of High-Fidelity Radar Cross Section Simulation Using OpenTURNS PCE, Th. Gricourt (ONERA)

When a radar wave illuminates an object, it reflects back; the energy that is reflected can be quantified by the quantitity known as the Radar Cross Section (RCS). Due to the geometry of irradiated objects, such as airplanes, this RCS can vary significantly depending on angle at which the object is seen by the radar. In a scenario where an airplane is approaching head-on within ±10° along the roll and pitch axes, the RCS cannot be approximated to a single value over this range. Therefore, a study on uncertainty propagation is necessary. To conduct this study, Polynomial Chaos Expansion (PCE) from the OpenTURNS library was utilized to accurately predict the RCS values based on angles calculated using a high-fidelity code: Maxwell3D.

otkerneldesign: an OpenTURNS module for kernel-based design of experiments, E. Fekhari (EDF R&D)

The construction of numerical designs of experiments is an essential part of uncertainty quantification, both used to build space-filling learning sets for surrogate models and to propagate uncertainties efficiently. Kernel-based sampling techniques have emerged in the machine learning community as a flexible and powerful alternative to the historical design of experiments methods natively implemented in OpenTURNS (e.g., Monte Carlo, quasi-Monte Carlo, or Latin hypercube sampling). For example, the so-called "kernel herding" algorithm relies on a dissimilarity measure (called "maximum mean discrepancy") to iteratively build a sample according to a target distribution or a target dataset. The Python module named "otkerneldesign" proposes a greedy implementation of this algorithm, using core OpenTURNS objects such as its covariance models. In this talk, we will first introduce the methodology of kernel herding and then present the otkerneldesign module and its documentation.

Benchmark of integrated solutions for uncertainty quantification, B. Kerleguer (CEA)

The need for uncertainty quantification in simulation code is already well established among openTURNS users. However, there are now a wide variety of tools available for carrying out these studies. The question of the ‘best’ tool therefore arises. The aim of this presentation is to take a look at these tools, in particular those with advanced HCI. Various elements of the ABCD method will be covered. The focus will be on physical problems or problems related to physical problems.