Session: Methods for Uncertainty Quantification, Sensitivity Analysis, and Prediction

Paper Number: 158645

158645 - Advanced Methods for Uncertainty Quantification in the Context of Composite Structures 

Abstract: 

Aerospace industry is increasingly relying on Modeling and Simulation (M&S) capabilitiesto support faster development of new aircrafts, embedding breakthrough technologies.  Certification issues and new Means of Compliance related to those capabilities have to be addressed [DOELAND2020]. Standards for Verification and Validation (V&V) of M&S [ASMEV&V10, ASMEV&V90] propose frameworks, processes and methods to assess the credibility of M&S capabilities with regards to target applications. Effective Verification and Validation (V&V) of M&S capabilities requires robust Uncertainty Quantification (UQ) methods, which, although complex, are essential for making fully informed decisions. In this context, it is of utmost importance to provide technical community (engineers, program managers, authorities) with comprehensive, efficient and robust end-to-end UQ approaches.

In the present work, focus is put on structural applications, in particular for composite aircraft structures which necessitate a very large number of tests to be conducted prior to certification and would benefit from more simulations. We consider M&S capabilities to predict the damage evolution of laminated composites that have been developed for several years by the academic community [LAURIN2007, MAIMI2007, RUIZ2022]. Due to the intrinsic complexity of such models covering multiple physical phenomena and their couplings, as well as their computational cost, uncertainty quantification methods should be chosen with care. Also, the industrial context leads to very limited experimental data. The following UQ issues have then quickly been raised when it has come to V&V process, from calibration to validation and exploitation of the M&S capabilities:

-        Although experts developing the models know how to determine the value of parameters from specific calibration experiments, could we define smarter and reduced sequence of tests to calibrate complex models while minimizing the uncertainties due to measurement errors?

-        Once robust calibration tests are defined, proper methods should be used to quantify the uncertainties on input quantities from limited number of experiments. Hence, how to apply Bayesian calibration to infer statistics on input quantities of such models?

-        Due to the large number of different load cases and configurations in the target application domain, only few experiments are available for calculating the validation metrics, while number of simulations is constrained as well. What is the uncertainty associated to the limited number of experimental and simulated samples used for validation?

-        Composite damage models are well suited for simulation at the scale of coupons and elements, which can be used to compute design allowable for higher scales. How to efficiently propagate uncertainties throughout the model to predict such quantities?

-        Relevant definition of lower level experiments for calibration and validation should be informed by the quantification of influent parameters at the higher levels. How to analyse sensitivity of larger structures to local material uncertain parameters?

 The present contribution will give an overview of the advanced methods that have been investigated to solve these issues in the framework of TRUST project, illustrated by applications on composite damage models and load cases relevant to composite aircraft structures.

Presenting Author: Ludovic Barriere IRT Saint Exupery

Presenting Author Biography: Research project leader
PhD in computational Mechanics

Authors:

Ludovic Barriere IRT Saint Exupery
Sébastien Bocquet IRT Saint Exupery
Clément Laboulfie IRT Saint Exupery
Marie Guerder Institut Clément Ader
Paolo Minigher AMADE University of Girona
Albert Turon AMADE University of Girona
Jean-Charles Passieux Institut Clément Ader
Jean-Philippe Navarro IRT Saint Exupery