Prediction of interlayer shear strength of double-layer asphalt using novel hybrid artificial intelligence models of ANFIS and metaheuristic optimizations

•Interlayer Shear Strength of Double-Layer Asphalt was predicted using optimized ANFIS models.•Temperature, normal pressure, and nominal maximum aggregate size were used as input variables.•R2, RMSE, MAE were used to validate and compare the models.•Results show that ANFIS-IWO model is slightly better than ANFIS-CA and ANFIS-DE models. The pavement damages such as slipping, rutting, top-down cracking of flexible pavement are very common during the operation process. These damages are mainly caused by interlayer shear stress acting between the asphalt layers during traffic movement. This is one of the main reasons for reduced pavement service life and increased maintenance costs. Therefore, the Interlayer Shear Strength (ISS) parameter needs to be evaluated and accurately predicted. Generally, ISS of double–layer asphalt concrete is measured directly in the field or on the samples in the laboratory, which involves much time and cost. Therefore, in this study, we have estimated ISS of asphalt pavement based on three input parameters: temperature, normal pressure and aggregate diameter using novel hybrid machine learning models namely Culture Algorithm based Adaptive Neuro Fuzzy Inference System (ANFIS-CA), Differential Evolution based Adaptive Neuro Fuzzy Inference System (ANFIS-DE) and Invasive Weed Optimization based Adaptive Neuro Fuzzy Inference System (ANFIS-IWO). Estimated shear strength was compared with the direct measured shear strength for the validation of model’s performance. In this study results of 180 double-layer asphalt samples which were fabricated by three different mixtures of Dmax 12.5, Dmax 19 and Dmax 25 and tested by shear device in the laboratory with five levels of normal pressure (0, 0.14, 0.2, 0.4, 0.6 MPa) at three temperature levels (25, 40 and 60 °C) were considered. Standard statistical measures namely Mean Absolute Error (MAE), Root-Mean-Square Error (RMSE), and Correlation Coefficient (R) were used to assess the models’ prediction capabilities. The results exhibit that all the suggested models have high accuracy of prediction, but performance of ANFIS-IWO (MSE = 0.02, MAE = 0.024, RMSE = 0.047, and R = 0.953) is slightly better than ANFIS-CA (MSE = 0.04, MAE = 0.031, RMSE = 0.047, and R = 0.966) and ANFIS-DE (MSE = 0.03, MAE = 0.034, RMSE = 0.057, and R = 0.953). Therefore, all the proposed models are suitable and powerful potential tools for the accurate prediction of the ISS for the proper consideration of pre-construction