Application of the intuitive k-NN Estimator for prediction of the Marshall Test (ASTM D1559) results for asphalt mixtures

This paper introduces a novel approach for modeling stability test data, based on the intuitive k-NN Estimator. In the proposed model, Marshall Briquettes’ features are explored and weighed using the genetic k-nearest neighbor approach, and then various distance metrics are applied for measuring the similarities among the features affecting the values of target parameters. The weighed features and obtained distance array are used to predict real values of flow, stability and Marshall Quotient. In experimental studies, the real measured data was used and the values of flow, stability and Marshall Quotient (MQ) were estimated. The test results have shown that weighted features have a primary role in the value prediction of target parameters. Moreover, the proposed model successfully explores the effects of different features on target parameters and predicts the real values of the parameters with a high accuracy rate. As far as this approach is concerned, useful and invaluable information is presented to the asphalt mixture definition with the proposed model. It is thought that a practical solution is possible with the GP method for understanding Marshall Test parameters, and in some measure, in context with the permanent deformation with the MQ method or flow comment.

► We studied a novel approach for modeling of stability test data based on the IKE. ► Marshall Briquettes’ features are explored and weighted by the genetic k-NN approach. ► Various distance metrics are used for measuring the similarities among the features affecting the values of target parameters. ► The weighed features and obtained distance array are used to estimate real test values. ► The proposed model predicted the real test values with a high accuracy rate.