Developing an optimized faulting prediction model in Jointed Plain Concrete Pavement using artificial neural networks and random forest methods

Predicting faulting failure is useful in the optimal concrete pavement design. In this study, artificial neural networks and the random forest method have been used to predict the amount of this failure. The general prediction model was created by inserting 32 available input variables into artificial neural networks. An integer two objectives optimisation problem was designed to select features that significantly affect the faulting. After applying this method, 19 important variables were identified and used to develop two simplified models based on artificial neural networks and the random forest method. It is shown that the simplified model developed by artificial neural networks is the best model to accurately predict the faulting considering the number of input variables. The cumulative number of days when the precipitation is more than 12.7 mm, the elastic modulus of concrete slab, the number of days passed since the pavement was built, base thickness, the cumulative ESALs in the traffic lane, and the annual average number of days when the temperature is more than 32 degrees C were identified as the most important parameters in predicting faulting using the random forest method. A Sensitivity analysis has been then performed on these variables and optimal values were determined.

Automated summary

Predicting faulting failure in concrete pavement design is achieved through artificial neural networks and random forest method, with an optimized model developed to identify key features influencing failure. Factors such as cumulative precipitation, concrete slab modulus, base thickness, and temperature play crucial roles in predicting faulting accurately. Sensitivity analysis helps determine optimal values for these key parameters.