Gypsum provides excellent fire protection because it dehydrates at a temperature around 120 °C. Dehydration is an endothermic chemical reaction absorbing energy and thus acting as a heat barrier. In the dehydration process, the water of crystallisation is released and transformed into vapour, which is transported through the porous gypsum material by pressure and by molecular diffusion. When migrating into colder regions, vapour can condense. Vapour transport and phase change are heat transport mechanisms which modify the temperature.
Several numerical simulations on the fire behaviour of gypsum boards have been published in the literature. The simplest models are pure heat conduction models, where the energy consumption due to dehydration is introduced as a heat sink or by an apparent heat capacity. Heat conduction models have been used for single boards [1] and [2] and for assemblies of two boards with a cavity [3], [4], [5], [6] and [7]. Heat conduction models use temperature-dependent material properties, namely enthalpy of dehydration, density, and thermal conductivity, which have to be determined experimentally. However, these parameters depend on the reaction temperatures, which in turn depend on the partial vapour pressure and the heating rate. In most models, the reaction temperatures are chosen at fixed temperatures, which are calibrated with fire tests. Such models give reasonable results for standard engineering applications. However, some details cannot be well captured, even with adjusted material parameters [8]. One of these details is the temperature plateau at 100 °C usually observed at the unexposed face of a gypsum board [2], [3], [4] and [6].
An obvious step for improving the heat transfer model is to include mass transfer. Since the dehydration produces water vapour, including the effect of moisture transport is a natural extension. Literature on moisture transport in gypsum is very scarce. Manzello et al. [9] describe a full model with a gas mixture consisting of air and vapour, transport by pressure and diffusion, and including condensation and evaporation. A kinetic model of Arrhenius-type is also included. Details are given by Kukuck [10] who provides equations and material parameters. Unfortunately, due to a wrong interpretation of the literature, the author used a much too low diffusion coefficient (see Table 3). Nevertheless, many useful links to material properties are provided. Simulations of a wall assembly with two boards show quite good agreement with measurements. Despite the advanced model, mass transfer had virtually no effect in this investigation. Craft et al. [11] also present a model with a gas mixture, but only consider transport by pressure, omitting molecular diffusion. However, their simulated pressure is unrealistically high due to a much too low permeability, as admitted by the authors. The model also includes condensation and evaporation and an Arrhenius-type kinetic model for the dehydration. The simulations agree remarkably well with experimental results, except that the temperature plateau is not well captured. Ang and Wang [12] and [13] adapted a model originally developed for concrete [14]. This model uses a single variable for the moisture content describing the liquid water in the pores. Vapour content and phase change are implicitly integrated in the constitutive equations. The constitutive equations are of semi-empirical nature and closely related to concrete. Whether an adaption to gypsum is feasible remains questionable. The simulations show that even by adjusting the permeability, the plateau cannot be reproduced. On the other hand, the authors were able to largely improve the pure heat conduction model by introducing a correction factor.
More recently, several authors investigated the influence of vapour transport and compared simulated temperatures to experimental data from Ghazi Wakili et al. [2]. As the present article also uses the same benchmark, the different models can easily be compared. Shepel et al. [15] present a heat and mass transfer model, where the gas is assumed to be pure vapour with pressure-driven transport. Valuable information are the permeability measurements also presented in this paper. The pure vapour model comes to its limits when condensation is included. Condensation reduces the vapour pressure, but the vapour pressure is always equal to the total pressure in this model. Despite this deficiency, a large improvement was found by including mass transfer. Kontogeorgos and Founti [16] describe a model with an air-vapour mixture transported by pressure and diffusion. However, condensation is neglected. Also here, the authors found a significant improvement when including vapour transport. According to this model, diffusion is the major transport mechanism for vapour. As will be shown below, the improvements found in the papers discussed in this paragraph are partially due to inconsistent formulations of the energy equation.
Models for heat and mass transfer in porous media have been developed also for other applications. In particular, the literature on concrete provides many relevant formulations [17] and [18]. Other references that proved useful relate to food processing [19], geomechanics [20], and chemical engineering [21]. General, more theoretical contributions are [22], [23] and [24].
The literature for heat and mass transfer in gypsum shows that many different models can lead to reasonable results. However, due to their complexity it is difficult to see, which physical mechanisms are important and what models are the most appropriate ones. In any case, the comparison with measured temperatures is not sufficient for a model assessment. In this paper we present a series of models with increasing complexity, based on a rigours mathematical formulation and carefully investigated material parameters. In this way, the relative importance of different mechanisms is more clearly visible. The base model is a heat conduction model without mass transfer, similar to many other models in the literature. Material parameters and temperatures measured during a fire test are taken from [2]. However, some material parameters are adjusted within justifiable limits until a partial fit with measurements is obtained. The next model includes vapour transport by pressure and diffusion, and the final model also considers condensation and evaporation. A main result is a consistent energy equation in the presence of phase change. Some formulations encountered in the literature are in error, and effects attributed to mass transfer are in reality due to an inconsistent energy equation. Not considered in this article are the kinetics of dehydration.