Effect of the water content measurements on the estimation and identifiability of water diffusion and convection mass transfer coefficients. (November 2015)

• Record Type: Journal Article
• Authors: Mercier, Samuel; Marcos, Bernard; Moresoli, Christine; Mondor, Martin; Villeneuve, Sébastien
• Is Part Of: International journal of heat and mass transfer. Volume 90(2015:Nov.)
• Page Start: 480
• Page End: 490
• Publication Date: 2015-11
• Subjects: Identifiability -- Uncertainty -- Mass transfer -- Diffusion coefficient -- Least-squares minimization; Heat -- Transmission -- Periodicals; Mass transfer -- Periodicals; Chaleur -- Transmission -- Périodiques; Transfert de masse -- Périodiques; Electronic journals; 621.4022
• DOI: 10.1016/j.ijheatmasstransfer.2015.06.082
• Languages: English
• ISSNs: 0017-9310

Abstract

Highlights: The identifiability of a mass transfer model was verified using the profile likelihoods. The mass transfer coefficients are not identifiable from noisy average water content. The coefficients are estimated more accurately from internal water content. Observations near the surface are more important for the coefficients estimation. Asymptotic confidence intervals do not accurately describe the uncertainty. Abstract: The aim of this work was to determine the impact of water content measurements on the structural and practical identifiability of the diffusion and the convection mass transfer coefficients. Global or internal water content were generated from the diffusion model with the addition of Gaussian noise. The diffusion and convection mass transfer coefficients were estimated from the water content by least-squares minimization and their confidence interval was calculated using the profile likelihood and the asymptotic methods. A unique combination of mass transfer coefficients reproduced exactly water content values without noise, indicating their structural identifiability. The coefficients were practically identifiable from the internal water content or the global water content if the noise intensity was less than 2% and 5%, respectively. The confidence intervals of the coefficients were asymmetric and were not accurately described by the asymptotic method.