Reducing Airfield Rigid Pavement Modeling Complexity: An Exercise Using the Principles of Similarity

Pavement design and evaluation analysis use mechanistic models to estimate pavement responses to applied loads. Finite element modeling is a common technique used to quickly and efficiently model rigid pavements that incorporate more complex phenomena that constructed, in-service slabs experience. While adding complexity increases the accuracy of the modeling when estimating pavement responses, significantly more computing effort is required. When combined with a cumulative-damage-based structural analysis, multiple model runs are needed to estimate damage over the number of incremental steps used. To bypass direct finite element modeling for multiple pavement systems, design methodologies such as the Mechanistic-Empirical Pavement Design Guide use artificial neural networks to store specific pavement response information for rapid recall as a type of non-linear regression made from pre-analyzed cases of a known set of input variables. These methodologies use the Principles of Similarity to reduce the complexity of modeling the pavement layering and environmental loads by considering a single reference pavement structure. Complexity can be reduced from 21 variables to nine key variables for modeling airfield pavements without introducing error and minimizing the total runs used. This paper provides a review of the Principles of Similarity and discusses how they are used to generate an efficient dataset for artificial neural network development. Examples showing how a single representative pavement system can yield proportional and scalable responses to numerous equivalent pavement systems are provided to illustrate the power of the Principles of Similarity in reducing modeling complexity and computational demands for higher-level pavement analysis efforts.