Bayesian inference for estimating thermal properties of a historic building wall. (September 2016)

• Record Type: Journal Article
• Authors: Berger, Julien; Orlande, Helcio R.B.; Mendes, Nathan; Guernouti, Sihem
• Is Part Of: Building and environment. Volume 106(2016)
• Page Start: 327
• Page End: 339
• Publication Date: 2016-09
• Subjects: Bayesian inference -- Identification problem -- Inverse heat transfer problem -- Real thermal performance -- Model calibration; Buildings -- Environmental engineering -- Periodicals; Building -- Research -- Periodicals; Constructions -- Technique de l'environnement -- Périodiques; Electronic journals; 696
• DOI: 10.1016/j.buildenv.2016.06.037
• Languages: English
• ISSNs: 0360-1323

Abstract

Abstract: In this paper, the use of Bayesian inference is explored for estimating both the thermal conductivity and the internal convective heat transfer coefficient of an old historic building wall. The room air temperature, as well as the temperatures at the surface and within the wall have been monitored during one year and then used to solve the identification problem. With Bayesian inference, the posterior distributions of the unknown parameters are explored based on their prior distributions and on the likelihood function that models the measurement errors. In this work, the Markov Chain Monte Carlo method is used to explore the posterior distribution. The error of the inadequacy of mathematical model are considered using the approximation error model. The distribution of the estimated parameters have a small standard deviation, which illustrates the accuracy of the method. The parameters have been compared to the standard values from the French thermal regulations. The heat flux at the internal surface has been calculated with the estimated parameters and the standard values. It is shown that the standard values underestimate the heat flux of an order by 10%. This study also illustrates the importance of the preliminary diagnosis of a building with the estimation of the thermal properties of the wall for model calibration. Highlights: An historic building wall has been monitored during one year. Bayesian inference is explored for estimating thermal properties. The thermal conductivity and internal convective transfer coefficient are estimated. The importance of preliminary diagnosis of a building for model calibration is shown.