Asphalt concrete healing is one of the important concepts related to flexible pavement structures. Fatigue endurance limit (FEL) is defined as the strain limit under which no damage will be accumulated in the pavement and is directly related to asphalt healing. Pavement section designed to handle a strain value equivalent to the endurance limit (EL) strain will be considered as a perpetual pavement. All four-point bending beam fatigue testing results from the NCHRP 944-A project were extracted and utilized in the development of artificial neural network (ANN) EL strain predictive model based on mixture volumetric properties and loading conditions. ANN model architecture, as well as the prediction process of the EL strain utilizing the generated model, were presented and explained. Furthermore, a stand-alone equation that predicts the EL strain value was extracted from the developed ANN model utilizing the eclectic approach. Moreover, the EL strain value was predicted utilizing the new equation and compared with the EL strain value predicted by other prediction models available in literature. A total of 705 beam fatigue lab test data points were utilized in model training and evaluation at ratios of 70%, 15%, and 15% for training, testing, and validation, respectively. The developed model is capable of predicting the EL strain value as a function of binder grade, temperature, air void content, asphalt content, SR, failure cycles number, and rest period. The reliability of the developed stand-alone equation and the ANN model was presented by reasonable coefficient of determination (R2) value and significance value (F).

One of the important distresses affecting the design of flexible pavement is fatigue cracking. It occurs because of repeated traffic loading cycles as a longitudinal or interconnected crack that was initiated as a bottom-up or top-down crack, depending on many factors that include, but not are limited to, high shear stresses and strains between or near truck tire edges as well as material related properties such as aging. For hot mix asphalt (HMA) of a thin layer, the fatigue crack appears in the outer vehicle wheel path, while for thick HMA layers the inner wheel path will crack; this phenomena is mainly because of different tire contact stress distribution between thin and thick pavements (1, 2).

The ability of HMA to retreat to its initial condition after being subjected to loading when enough time is provided between two successive loading cycles is defined as asphalt healing (3). Asphalt healing is the cause of the endurance limit (EL) strain concept, which is defined as the strain value under which no damage will accumulate in the HMA layer; therefore, the resulting pavement section will be a perpetual pavement (4).

There is a need to have a reliable EL strain prediction model to be incorporated into the current mechanistic design approach. The current design approach assumes that a portion of the pavement sections’ fatigue life is consumed along with each loading cycle the pavement is subjected to. However, recent studies discussed that a well-built pavement section will not fail under fatigue cracking despite being subjected to many loading cycles; that is, the damage is not accumulating in the pavement section (5–7).

Artificial neural network (ANN) is being utilized by a growing number of researchers as a data management tool because of its wide-range capabilities in prediction and categorization. A variety of problems can be solved utilizing ANN modeling, such as problems related to function approximation and fitting as well as pattern recognition tasks (8). Thus, the EL strain prediction model presented within this research paper was developed utilizing the above-mentioned ANN model development technique.