Prediction of the High-Temperature Performance of a Geopolymer Modified Asphalt Binder using Artificial Neural Networks

Abstract

Complexity in the behaviour of an asphalt binder is further escalated with geopolymer (fly ash and alkali liquid) modification, thus making it difficult to accurately predict the performance of the binder. This study employs artificial neural network modelling to predict the complex shear modulus, storage modulus, loss modulus and phase angle outcomes of experimental results from dynamic shear rheometer (DSR) oscillation tests under four separate scenarios. The proposed artificial neural network models received test conditions (temperature and frequency) and three different geopolymer concentrations (3%, 5% and 7% by the weight of bitumen) as the predictor parameters. The variants of the optimal algorithms were Levenberg-Marquardt (LM), Scaled conjugate gradient and Polak-Ribiere conjugate gradient (CPG) training algorithms with different combinations of network structures and tan-sig and log-sig as activation functions. The coefficient of determination, covariance and root mean square error (RMSE) were used as statistical measures of model prediction performance. Based on the statistical performance indicators, the LM algorithm with a 3-5-1 network architecture and tan-sig as the activation function was the best performing model for predicting the complex modulus with R 2 values of 0.996 for the training dataset and 0.971 for the testing dataset and RMSE values of 0.118 and 0.139 for the training and testing datasets, respectively. Furthermore, it was observed that the least efficient model was the phase angle prediction model developed with the CPG training algorithm, which had a 3-8-1 network architecture and log-sig as the activation function. The model yielded R 2 values of 0.909 and 0.829 for the training and testing datasets, respectively. Poor prediction performance for the testing dataset indicated that the model was unable to learn complexity in the data and would perform below a significance level of 0.90 in predicting using untrained data.

A Hybrid Neural Network for Predictive Model in A Plastic Injection Molding Process

“…The maximum number of iterations is limited Hence, the HNN (9-7-1) with seven hidden neurons giving the lowest RMS in the study for %volume shrinkage prediction model was chosen and the RMS error of testing data for such a configuration was 4.87%, while HNN (9-3-1) with three hidden neurons was chosen for warpage prediction model with the lowest RMS of 2.47%. For backpropagation neural network (BPNN), since it is a well-established neural network, the detailed algorithm is not discussed in this work, but appropriate references are provided in [16][17][18]. The lowest RMS error of 5.01% and 2.41% was obtained for %volume shrinkage and warpage when using the BPNN (9-8-1) and BPNN (9-6-1), respectively.…”

Emerging Technologies of Deep Learning Models Development for Pavement Temperature Prediction

“…The sigmoid function calculates the exportation of the cell. A thin component selects the cell state and multiplies it with the sigmoid function's output as shown in equations ( 7) - (8).…”

“…Researchers have developed regression models for predicting asphalt temperature by measuring pavement temperature using solar radiation and air temperature (AirT) [2][3][4][5][6][7][8][9][10]. These models produced false data for continuous temperature variations because they did not consider the environmental factors that may influence the surface temperature prediction, such as wind speed, wind direction, and relative humidity (RH).…”

Stiffness Modulus and Marshall Parameters of Hot Mix Asphalts: Laboratory Data Modeling by Artificial Neural Networks Characterized by Cross-Validation

“…In general, the closer the value of R is to the unity, the stronger the results of the linear relation between t and y, thus confirming that the training has been completed successfully (if R 1 for the training data set) and that the degree of generalization achieved can be considered optimal (if R 1 for the testing data set). The mean squared error and the correlation coefficient have already been used in previous performance analysis of some ANNs designed to predict the mechanical parameters of HMA mixtures [65,66,[72][73][74][75][76]90].…”