Westergaard Curling Solution Reconsidered

An in-depth and systematic examination is presented of the effect of temperature gradients on slab-on-grade pavements. The main objective for the examination was the development of practical design tools for use in a typical engineering office. This has been achieved by a critical reconsideration of the literature, a synthesis of currently available analytical resources, and the implementation of recent technological achievements promulgated in related areas of engineering. Prominent among these are the application of the principles of dimensional analysis, the finite element method, advanced statistical regression analysis, and artificial neural networks (ANN). A number of ANNs have been trained for the curling problem, and in several instances they are found to be more efficient predictive tools than corresponding statistical regression equations. It is found that the most important shortcomings of the Westergaard curling solution are the assumption of continuous contact between slab and subgrade (infinite slab self-weight) and the explicit treatment of only daytime conditions. Although Westergaard’s curling-only predictions are significantly inferior to those from ANN and statistics, his load-plus-curling predictions exhibit approximately the same scatter as those from these two more modern and nominally more sophisticated tools. The case of Westergaard’s curling solution can serve as an example pointing to the usefulness and desirability of theoretical solutions, even when these are achievable only on the basis of considerable abstraction and simplification.