Non-destructive thermal diagnosis is necessary for identification and quantification of structural defects and verification of construction performances. In this paper, we test different inverse heat transfer models to identify the thermal conductivity of a building's wall in its environment. The approaches proposed and used in this paper rely only on non-intrusive measurements: inside and outside wall surface temperatures and heat flow through the wall. First, a Bayesian statistical dynamic inference method, which has the advantage to quantify the unknown parameter and its credible interval, is presented. This method considers the uncertainties of the measured temperature and heat flow data and of the unknown thermal properties. Markov chain Monte Carlo (MCMC) algorithm is used to explore the posterior distribution. Then, the average and the dynamic procedures ISO 9869 (I. 9869-1, 2014) are introduced. Finally, the probabilistic distributions of the unknown parameters are presented and compared to the standard results. The impact of experimental conditions (average indoor-to-outdoor temperature) and the measurements length on the accuracy of the results are discussed. The relationship between the number of iterations of the MCMC, time series length, shape of the prior distribution and accuracy are studied as well as the simulation time to run the inverse models. The Bayesian approach gives the most accurate results and has the advantage of considering several unknowns (conductivity and volumetric heat capacity), which is not the case for the studied standard. The Bayesian method needs much shorter time series than the ISO standard and produces robust results at all times of year, including when the average indoor-to-outdoor temperature difference was low. (C) 2018 Elsevier Ltd. All rights reserved.
Page:
574---593
Related
Batch download
Cited By
noting
Similar Literature
Submit Feedback
Please wait while the file you selected is being converted