ADAPTIVE NEURO FUZZY INFERENCE SYSTEM FOR PREDICTING FLEXIBLE PAVEMENT DISTRESS IN TROPICAL REGIONS

The construction of asphalt pavement road networks is very costly for the government budget. A flexible pavement maintenance prediction is unavoidable to reduce government expenses and increase the service life span of pavements. This paper aims to predict flexible pavement maintenance in tropical regions using an Adaptive Neuro-Fuzzy Inference System (ANFIS) model. This research proposes Artificial Intelligent (AI) method with the statistical approach. Moreover, three ANFIS models with different membership functions were used to make predictions. The data from two types of flexible pavements, single distress and combined distress, were used to make predictions. Four parameters, i.e., severity, density, road function, and treatment technique, were used as inputs, while the target parameter is the pavement condition. To obtain the Fuzzy Inference System (FIS) for the three ANFIS models, three different structures, i.e. (3 3 3 3), (3 3 2 2) and (3 2 3 2) as well as three rules 81, 36 and 36, respectively, were used to train the models with 30 epochs. The ANFIS model's finding among the three has a regression square (R-2) of 0.934 and Root Mean Square Error (RMSE) of 0.565. Results show that the predictions made using ANFIS are accurate and closely approximate the target data.