Parameter estimation of thermal response test: A Bayesian inference using Markov chain Monte Carlo - Metropolis Hastings (MCMC-MH) approach

Abstract

To understand and model the thermal response of a body engulfed in fire, the knowledge of the geometrical and thermo-physical properties is necessary. In this work, simultaneous estimation of parameters of convective heat transfer coefficient (h) and configuration factor (F) of a thermal response test is accomplished with reported fire experiments. The measurements of these parameters in a fire environment is complex. Sometimes it demands numerical simulations and semi-empirical modelling. Therefore, a Bayesian driven Markov Chain Monte Carlo Metropolis-Hastings (MCMC-MH) approach is applied as an alternative to estimate the unknown parameters h and F simultaneously. The priors for h and F are generated using the offline Bayesian method. The generated priors are given as means of the Gaussian distribution and parameters are estimated simultaneously by generating the samples dynamically. The uncertainty of the estimates is reported in the form of their standard deviations. Moreover, the validation of the estimated parameters is performed. It is observed that the simulated temperature distribution with estimated parameters is in good agreement with the measured temperature distribution as their deviation lies at less than 1%. The efficacy of the output of the Bayesian inference framework is also reported.