No full text
Unpublished conference/Abstract (Scientific congresses and symposiums)
Bayesian parameter inference for PICA devolatilization pyrolysis at high heating rates
Coheur, Joffrey; Arnst, Maarten; Magin, Thierry et al.
20193rd International Conference on Uncertainty Quantification in Computational Sciences and Engineering
 

Files


Full Text
No document available.
Annexes
coheur_presentation_UNCECOMP2019.pdf
Publisher postprint (9.67 MB)
Download

All documents in ORBi are protected by a user license.

Send to



Details



Keywords :
Inverse problem; Bayesian inference; Markov chain Monte Carlo; Chemical Kinetics; Pyrolysis
Abstract :
[en] The heat shield of high speed reentry spacecraft is often made up of porous ablative thermal protection materials (TPMs) that can accommodate high heating rates and heat loads through phase change and mass loss. When the temperature increases, those materials absorb heat and start to pyrolyze, releasing gases that interact with the surrounding flow. Modeling the species production and the material decomposition rate is important for their use in numerical simulations for the robust determination of heat shield thickness. To this end, pyrolysis experiments have been performed on TPMs in order to determine the kinetic parameters of chemical laws that govern mass loss and species production rates [Wong et al., Polym. Degrad. Stabil., 112:122–131, 2015; Bessire and Minton, ACS Appl. Mater. Interfaces, 9:21422—21437, 2017]. Samples are heated in a furnace and the mass loss is measured while the species produced are collected. In the present work, we investigate parameter calibration of law governing the thermal decomposition of TPMs in a Bayesian inference framework using experimental data of the evolution of gas species production over time. We discuss the problem of highly correlated parameters and we propose a reparametrization of the equations that lead to a scaled posterior density function which is more efficiently sampled using a multivariate Gaussian distribution in the random walk Metropolis-Hastings algorithm. We study the influence of reducing the available data set in order to avoid over-confidence in the parameter values. To that aim, we propose to extract important features based on zero-gradient and inflection production rate temperatures or temperature of reaction. Finally, measurement noise and model error norm appear to be critical in the inference results. Measurement noise tending to zero can lead to regions in the data set that are over-constrained and we propose different strategies in order to relax the problem in regions of smaller significance and we apply several norm for the model error definition.
Disciplines :
Aerospace & aeronautics engineering
Author, co-author :
Coheur, Joffrey  ;  Université de Liège - ULiège > Département d'aérospatiale et mécanique > Computational and stochastic modeling
Arnst, Maarten ;  Université de Liège - ULiège > Département d'aérospatiale et mécanique > Computational and stochastic modeling
Magin, Thierry;  von Karman Institute for Fluid Dynamics > Aeronautics and Aerospace Department
Chatelain, Philippe;  Ecole Polytechnique de Louvain > Institute of Mechanics, Materials and Civil Engineering (IMMC) > Thermodynamics and fluid mechanics
Language :
English
Title :
Bayesian parameter inference for PICA devolatilization pyrolysis at high heating rates
Publication date :
24 June 2019
Event name :
3rd International Conference on Uncertainty Quantification in Computational Sciences and Engineering
Event organizer :
Eccomas
Event place :
Hersonissos, Crete, Greece
Event date :
du 24 juin 2019 au 26 juin 2019
Audience :
International
References of the abstract :
https://2019.uncecomp.org/files/uploads/general/compdyn_2019_programme_online.pdf
Funders :
Fédération Wallonie Bruxelles. Fonds de la Recherche Scientifique - F.R.S.-FNRS
Available on ORBi :
since 12 July 2019

Statistics


Number of views
99 (17 by ULiège)
Number of downloads
56 (10 by ULiège)

Bibliography


Similar publications



Contact ORBi