Investigation of the fatigue life relationship among different geometry combinations of the 3-point bending cylinder (3PBC) fatigue test for asphalt concrete

Laboratory testing is widely used to characterise fatigue properties of asphalt mixtures for the design of asphalt pavements. The recently introduced Three-Point Bending Cylinder (3PBC) test has proven to be an advantageous laboratory testing alternative. However, the standard sample size for the 3PBC test may hamper its use for testing as-built thin lifts of asphalt pavement. In this study, the potential to alter the sample size for 3PBC test was investigated. Various numerical techniques were explored to relate the fatigue performance in terms of the number of cycles to failure (N-f) among varying specimen diameters. In addition to N-f of the tested 3PBC sample diameter N-f (D-i), strain level; 3PBC test sample diameter D-i; and dynamic modulus |E*|; were selected as input variables to correlate with the N-f for the reference geometry N-f(D-.R.). Initially, multi-linear regression (MLR) analysis was applied to inspect if any cogent relation could be achieved among the N-f values for different 3PBC sample diameters. In addition, multigene genetic programming (MGGP) and artificial neural network (ANN) analyses were performed to check if a better fit could be achieved. Both these techniques, however, revealed similar statistical results to the MLR model, with the ANN model achieving higher statistics.