A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Evaluation of Pavement Condition Index Using Artificial Neural Network Approach
https://orcid.org/http://orcid.org/0000-0003-1772-9861
The soft computing technique, such as an artificial neural network (ANN) is a modelling tool that is used to predict the pavement condition index value. ANN is used to model highly variable non-linear distress available in the pavement. In this paper, the ANN model is trained and tested using three different algorithm Levenberg–Marquardt (LM), Bayesian regularization (BR) and scaled conjugate gradient (SCG) algorithm available in the NN toolbox of MATLAB-2015 version. The flexible pavement distress data collected from the field surveying are used to train and test the ANN model. The algorithms are trained and tested using the 400 data points gathered after the rainy season from four different National Highways (NH) of Bihar, India. The performance of ANN algorithms are determined using the distress density and the calculated pavement condition index (PCI) value. The prediction accuracy of the LM, BR and SCG algorithm was compared using the minimum statistical errors (MSE) and the coefficient of correlation. The LM algorithm has performed better than BR and SCG algorithms. LM algorithm has shown the highest correlation value (89%) compared to BR (76%) and SCG (58%) during training, testing and validation of data points. The models predicted the PCI value with a high correlation coefficient and low MSE value. The five-fold cross-validation has been performed to explain the accuracy of the ANN model which is higher compare to random forest and support vector machine (SVM). The ANN model has shown the highest model accuracy (73%) compared to the SVM model (72%) and RF model (61%).
Evaluation of Pavement Condition Index Using Artificial Neural Network Approach
https://orcid.org/http://orcid.org/0000-0003-1772-9861
The soft computing technique, such as an artificial neural network (ANN) is a modelling tool that is used to predict the pavement condition index value. ANN is used to model highly variable non-linear distress available in the pavement. In this paper, the ANN model is trained and tested using three different algorithm Levenberg–Marquardt (LM), Bayesian regularization (BR) and scaled conjugate gradient (SCG) algorithm available in the NN toolbox of MATLAB-2015 version. The flexible pavement distress data collected from the field surveying are used to train and test the ANN model. The algorithms are trained and tested using the 400 data points gathered after the rainy season from four different National Highways (NH) of Bihar, India. The performance of ANN algorithms are determined using the distress density and the calculated pavement condition index (PCI) value. The prediction accuracy of the LM, BR and SCG algorithm was compared using the minimum statistical errors (MSE) and the coefficient of correlation. The LM algorithm has performed better than BR and SCG algorithms. LM algorithm has shown the highest correlation value (89%) compared to BR (76%) and SCG (58%) during training, testing and validation of data points. The models predicted the PCI value with a high correlation coefficient and low MSE value. The five-fold cross-validation has been performed to explain the accuracy of the ANN model which is higher compare to random forest and support vector machine (SVM). The ANN model has shown the highest model accuracy (73%) compared to the SVM model (72%) and RF model (61%).
Evaluation of Pavement Condition Index Using Artificial Neural Network Approach
https://orcid.org/http://orcid.org/0000-0003-1772-9861
The soft computing technique, such as an artificial neural network (ANN) is a modelling tool that is used to predict the pavement condition index value. ANN is used to model highly variable non-linear distress available in the pavement. In this paper, the ANN model is trained and tested using three different algorithm Levenberg–Marquardt (LM), Bayesian regularization (BR) and scaled conjugate gradient (SCG) algorithm available in the NN toolbox of MATLAB-2015 version. The flexible pavement distress data collected from the field surveying are used to train and test the ANN model. The algorithms are trained and tested using the 400 data points gathered after the rainy season from four different National Highways (NH) of Bihar, India. The performance of ANN algorithms are determined using the distress density and the calculated pavement condition index (PCI) value. The prediction accuracy of the LM, BR and SCG algorithm was compared using the minimum statistical errors (MSE) and the coefficient of correlation. The LM algorithm has performed better than BR and SCG algorithms. LM algorithm has shown the highest correlation value (89%) compared to BR (76%) and SCG (58%) during training, testing and validation of data points. The models predicted the PCI value with a high correlation coefficient and low MSE value. The five-fold cross-validation has been performed to explain the accuracy of the ANN model which is higher compare to random forest and support vector machine (SVM). The ANN model has shown the highest model accuracy (73%) compared to the SVM model (72%) and RF model (61%).
Evaluation of Pavement Condition Index Using Artificial Neural Network Approach
Transp. in Dev. Econ.
Kumar, Rajnish (author) / Suman, Sanjeev Kumar (author) / Prakash, Gautam (author)
2021-10-01
Article (Journal)
Electronic Resource
English
Simulating pavement structural condition using artificial neural networks
| Taylor & Francis Verlag | 2016
Simulating pavement structural condition using artificial neural networks
| Online Contents | 2015
Simulating pavement structural condition using artificial neural networks
| Online Contents | 2016
Pavement friction management – artificial neural network approach
| Online Contents | 2016
Pavement friction management – artificial neural network approach
| Taylor & Francis Verlag | 2019