Influence of thermal boundary conditions and temperature distribution in concrete on frost scaling – experimental and numerical study

Frost resistance of concrete is a durability concern in cold climates such as Denmark. Though it was discovered decades ago that frost action causes damage in concrete, the fundamental mechanism of frost damage is still not completely understood. The aim of this PhD thesis is to improve the understanding of how frost deterioration evolves in concrete. It is attempted to achieve this aim by combining experimental studies and numerical modelling. Several studies in the literature have pointed to that thermal boundary conditions influence the temperature distribution in the concrete, and variation in temperature distribution may influence the extent of frost scaling. It is noted that the effect of thermal boundary conditions on frost scaling is rarely studied, and thus it remains a somewhat overlooked aspect of frost action.
This thesis focuses on the influence of thermal boundary conditions and resulting temperature distributions on frost scaling. In standardized testing, a liquid layer of a certain thickness is involved when concrete is exposed to freeze/thaw cycles. It is noted that the primary thermal boundary conditions that influence the temperature distribution in concrete are related to this layer and the temperature of the surroundings. Thus, the salt concentration in the external liquid, the thickness of the external liquid layer, and the surrounding temperature of the specimen being tested are considered thermal boundary conditions that influence the temperature distribution.
In the experimental part of the project, experiments are designed according to the principles of factorial design, where each thermal boundary condition is considered at two levels. The frost scaling is collected after seven freeze/thaw cycles, while the temperature distribution is recorded using type T thermocouples in these cycles. The statistical analysis of results shows that the salt concentration significantly affects frost scaling, which is not surprising. The interesting aspect is that the salt concentration is also interacting with other thermal boundary conditions such as the surrounding temperature. The physical interpretation of the statistical results shows that the extent of moisture transport can be related to the temperature distributions. It is concluded that both the unfrozen external liquid (above concrete) and unfrozen pore liquid (below the concrete surface) can be transported to the freezing area in the vicinity of the concrete surface. Thermal boundary conditions influence this transport and thus the extent of frost scaling.
Before developing a numerical model that mimics the experiments, the available models simulating frost action in cement-based materials are reviewed. The overview of the reviewed models provides valuable insights into the choices made by the model builders and illuminates some factors that need more attention in future studies regarding modeling frost action in cement-based materials. For example, freezing of supercooled liquid and moisture exchange with surroundings are known to be important in experimental research but have been treated very sporadically in numerical modeling. The review of the numerical models can help to decide a modelling framework, and important factors need to be considered in developing a model for specific objectives.
A numerical model is developed to simulate the temperature distribution in mortar considering boundary conditions similar to those considered in the experimental investigations. It is highlighted in the review of the numerical models that liquid supercooling is not given the necessary attention in numerical models. Therefore, special attention is given to supercooling and its effects on the freezing process. The model is developed using COMSOL Multiphysics. The freezing of the supercooled external liquid is modeled using an artificial heat source in the COMSOL, whereas the rest of the freezing process is modeled using modifications in the thermal properties of the external liquid and underlying mortar. Thermal properties are determined based on the volume fractions of ice and unfrozen liquid. Comparison of simulated and experimental results shows good agreement. However, there are some discrepancies, primarily due to the approach adopted to apply freeze/thaw cycles in the model. It is concluded that the model can capture the general effect of thermal boundary conditions to predict temperature distributions. The model predictions can be used to relate the extent of moisture transport with the temperature distributions and thus the extent of frost scaling. The model can also provide further insights to improve the understanding of frost action in concrete.