Artificial Intelligence Based Mix Design of Pavement Mixes: Fid-Bau Portal

Artificial Intelligence Based Mix Design of Pavement Mixes

Mallick, Rajib B. / Nivedya, M. K. / Veeraragavan, Ramkumar

This paper presents a framework for using an artificial intelligence (AI) based computational approach for accelerating pavement mix design. The experimentation work in pavement engineering can be optimized significantly if the designer can start the work with materials or mixes that are known to be good performers, with a certain degree of accuracy. Unfortunately, the state of the art in pavement engineering has not yet reached a point where the properties that are known to be correlated to pavement performance can be computed from first principles. However, a variation of the approach mentioned above can still be adopted, as long as there is a set of known material properties and their performance. The predictors can be isolated and then used for building classification models, and the model with the highest accuracy can be selected. Once the model is verified it can be inverted to predict which combinations of properties are desirable. These identified sets of properties will then become the basis for the starting of experimentation in the mix design procedure. The suggested framework is illustrated with an example of designing HMA mixes for adequate resistance against moisture damage.