Inferring and Propagating Kinetic Parameter Uncertainty for Condensed Phase Burning Models

详细信息    查看全文

• 作者：Morgan C. Bruns
• 关键词：Pyrolysis ; Kinetics ; TGA ; Bayesian inference ; Markov Chain Monte Carlo ; Uncertainty propagation ; FDS
• 刊名：Fire Technology
• 出版年：2016
• 出版时间：January 2016
• 年：2016
• 卷：52
• 期：1
• 页码：93-120
• 全文大小：31,779 KB
• 参考文献：1.McGrattan K, Hostikka S, McDermott R, Floyd J, Weinschenk C, Overholt K (2014) Fire dynamics simulator, technical reference guide volume 3: validation. Technical report, NIST
  2.McGrattan K, Hostikka S, McDermott R, Floyd J, Weinschenk C, Overholt K (2014) Fire dynamics simulator, technical reference guide volume 1: mathematical model. Technical report, NIST
  3.Lautenberger C, Fernanez-Pello C (2009) Generalized pyrolysis model for combustible solids. Fire Saf J 44:819–839CrossRef
  4.Stoliarov Stanislav I, Lyon Richard E (2008) Thermo-kinetic model of burning. Technical Report DOT/FAA/AR-TN08/17, FAA
  5.Lautenberger C, Rein G, Fernanez-Pello C (2006) The application of a genetic algorithm to estimate material properties for fire modeling from bench-scale fire test data. Fire Saf J 41:204–214CrossRef
  6.Chaos M, Khan MM, Krishnamoorthy N, de Ris JL, Dorofeev SB (2011) Evaluation of optimization schemes and determination of solid fuel properties for CFD fire models using bench-scale pyrolysis tests. Proc Combust Inst 33:2599–2606CrossRef
  7.Lautenberger C, Fernandez-Pello C (2011) Optimization algorithms for material pyrolysis property estimation. In: Fire safety ccience: 10th international symposium, College Park
  8.Rein G, Lautenberger C, Fernandez-Pello C, Torero JL, Urban D (2006) Application of genetic algorithms and thermogravimetry to determine the kinetics of polyurethane foam in smoldering combustion. Combus Flame 146:95–108CrossRef
  9.Bruns MC, Koo JH, Ezekoye OA (2009) Population-based models of thermoplastic degradation: using optimization to determine model parameters. Polym Degrad Stab 94(6):1013–1022CrossRef
  10.Gelman A, Carlin JB, Stern HS, Rubin DB (2003) Bayesian data analysis, 2nd edn. Chapman & Hall/CRC, Boca Raton
  11.Sivia DS (2006) Data analysis, 2nd edn. Oxford University Press, OxfordMATH
  12.Moser RD, Terejanu G, Oliver TA, Simmons CS (2012) Validating the prediction of unobserved quantities. Technical Report 12–32. The Institute for Computational Engineering Sciences
  13.Overholt KJ, Ezekoye OA (2014) Quantitative testing of fire scenario hypotheses: a bayesian inference approach. Fire Technol 1–33
  14.Overholt KJ (2013) Forward and inverse modeling of fire physics towards fire scene reconstructions. PhD thesis, The University of Texas at Austin
  15.Stoliarov SI, Safranova N, Lyon RE (2009) The effect of variation in polymer properties on the rate of burning. Fire Mater 33:257–271CrossRef
  16.Chaos M (2013) Application of sensitivity analyses to condensed-phase pyrolysis modeling. Fire Saf J 61:254–264CrossRef
  17.Bal N, Rein G (2013) Relevant model complexity for non-charring polymer pyrolysis. Fire Saf J 61:36–44CrossRef
  18.Schwarz G (1978) Estimating the dimension of a model. Ann Stat 6:461–464CrossRef MATH
  19.Bolstad WM (2010) Understanding computational Bayesian statistics. Wiley, HobokenMATH
  20.Metropolis N, Rosenbluth AW, Rosenbluth MN, Teller AH (1953) Equation of state calculations for fast computing machines. J Chem Phys 21:1087–1092CrossRef
  21.Hastings WK (1970) Monte carlo sampling methods using markov chains and their applications. Biometrika 57:97–109CrossRef MATH
  22.Haario H, Saksman E, Tamminen J (2001) An adaptive metropolis algorithm. Bernoulli 7:223–242CrossRef MathSciNet MATH
  23.Patil A, Huard D, Fonnesbeck CJ (2010) PyMC: Bayesian stochastic modelling in python. J Stat Softw 35:1–81CrossRef
  24.Stoliarov SI, Crowley S, Lyon RE, Linteris GT (2009) Prediction of the burning rates of non-charring polymers. Combust Flame 156:1068–1083CrossRef
  25.Stoliarov SI, Crowley S, Walters RN, Lyon RE (2010) Prediction of the burning rates of charring polymers. Combust Flame 157:2024–2034CrossRef
  26.Lyon RE, Safronava N, Senese J, Stoliarov SI (2012) Thermokinetic model of sample response in nonisothermal analysis. Thermochim Acta 545:82–89CrossRef
  27.Scott DW (1992) Multivariate density estimation: theory, practice, and visualization. Wiley, New York, pp 125–190MATH
  28.Ceamanos J, Mastral JF, Millera A, Aldea ME (2002) Kinetics of pyrolysis of high density polyethylene. comparison of isothermal and dynamic experiments. J Anal Appl Pyrolysis 65:93–110CrossRef
  29.McGrattan K, Hostikka S, McDermott R, Floyd J, Weinschenk C, Overholt K (2014) Fire dynamics simulator, user’s guide. Technical report, NIST
  30.Austin PJ, Buch RR, Kashiwagi T (1998) Gasification of silicone fluids under esternal thermal radiation part I. gasification rate and global heat of gasification. Fire Mater 22:221–237CrossRef
  31.American Society for Testing and Materials, West Conshohocken, Pennsylvania (2007) ASTM E 1354–04a, Standard test method for heat and visible smoke release rates for materials and products using an oxygen combustion calorimeter
  
• 作者单位：Morgan C. Bruns (1)
  
  1. National Institute of Standards and Technology, 100 Bureau Drive, Mail Stop 8665, Gaithersburg, MD, 20899, USA
  
• 刊物类别：Engineering
• 刊物主题：Civil Engineering
  Mechanics
  Characterization and Evaluation Materials
  Physics
  
• 出版者：Springer Netherlands
• ISSN：1572-8099

文摘

Kinetic parameters for serial pyrolysis reactions were calibrated from thermogravimetric analysis (TGA) data using Bayesian inference via Markov Chain Monte Carlo (MCMC) simulations assuming a serial reaction mechanism. Calibrations were performed for high-impact polystyrene (HIPS), bisphenol-A polycarbonate (PC), and poly(vinyl chloride) (PVC) at heating rates of 3 K/min and 10 K/min. The resulting parameter inferences are probabilistic as opposed to the point estimates calibrated in previous studies and are visualized using posterior probability density functions (PDFs) generated by kernel density estimation (KDE). Correlations between the parameters are identified and discussed. In particular, it is clear that pre-exponential constants and activation energies for a given reaction have a strong positive correlation. It is hypothesized that the degree of overlap in the posterior PDFs might be a measure of model adequacy. Point-estimates of the kinetic parameters were made by finding the mode of the posterior PDFs. For HIPS, it was determined that a one-reaction pyrolysis model is most appropriate, and that the posterior modes for \(\log \left( A_1\right) \) and \(E_1\) are \(19.5\,\log (1/s)\) and 292 kJ/mol, respectively, for the 3 K/min data. To evaluate the effect of kinetic parameter uncertainty on predictions of burning rate, samples from the posterior PDF were used to simulate gasification and cone calorimetry experiments using the fire dynamics simulator (FDS). In some cases, it was found that models with fewer parameters provided better predictions due to over-fitting associated with greater model complexity. Another important observation is that for the predictions of PVC cone calorimetry, the time to peak heat release rate can range from around 40 s to 180 s for a number of different kinetic parameter combinations that all fit the TGA data fairly well. It is argued that the proposed methodology is necessary for progress in modeling of condensed phase physics for fire problems as it supports both model validation and engineering predictions. Keywords Pyrolysis Kinetics TGA Bayesian inference Markov Chain Monte Carlo Uncertainty propagation FDS