Advanced statistical modeling of loading waveform, frequency, temperature, and SBS modification effects on asphalt concrete resilient modulus

Resilient modulus (Mr) is one the most important parameter for designing flexible pavements based on different damage criteria. The resilient modulus test measures the recoverable elastic strains of asphalt concrete under repeated loading, and therefore provides a stiffness-related parameter which is affected by temperature and loading conditions. The Mr of asphalt concretes is typically measured at specific temperatures and loading frequencies because of the high cost and time-consuming nature of testing; therefore, predictive models are necessary to predict the Mr at other temperatures and loading frequencies. The aim of this study is to present a framework for modeling and predicting the Mr of asphalt concretes by considering four parameters: loading waveform, loading frequency, temperature, and SBS modification. For this purpose, three advanced statistical techniques including Gene Expression Programming (GEP), Response Surface Methodology (RSM), and Hybrid Multivariate Adaptive Regression Spline-Pelican Optimization Algorithm (MARS-POA) have been employed. SBS, loading waveform, temperature, and loading frequency were used as the input parameters. The statistical comparison of developed models reveals that the MARS-POA method outperformed the others. The R 2 values for training and testing data were 0.9973 and 0.9732, respectively. Temperature has the greatest relative importance, while waveform has the least influence on Mr. In addition, master curves corresponding to waveform type and to SBS-modified or control mixture were calibrated in order to compare the results of MARS-POA model with those of master curves.