International Roughness Index Modeling utilizing Adaptive Neuro-Fuzzy Inference System (ANFIS)

As an essential component of pavement management systems (PMS), deterioration models have been adopted for predicting future conditions of pavement sections. It assists in selecting the best maintenance, repair, and rehabilitation decisions. As a result of the formation and growth of distresses such as cracks and rutting, pavement deterioration reduces serviceability and results in the failure of pavement sections. This paper aimed to develop a pavement condition forecasting model based on a pavement performance indicator called the international roughness index (IRI). The Long-term pavement performance (LTPP) database was used to obtain historical IRI data for pavement sections, including pavement sections in the United States (US) and Canada. Additionally, other pavement characteristics were collected to be used as model inputs, such as pavement age, annual precipitation, annual temperature, Freezing Index, Freeze Thaw, annual Average Humidity min, annual Average Humidity max, wind speed, the ratio of AADTT to AADT, KESAL, and SN. ANOVA analysis was implemented to select the most significant factors to include in the model. Moreover, Adaptive Neuro-Fuzzy Inference System (ANFIS) method was applied to develop the deterioration model. In preparation for model development, the dataset was cleansed and preprocessed. In addition, the data was split into 85% training and 15% testing. The ANFIS model analysis was performed using MATLAB software. The developed model's performance was measured using mean squared error (MSE), root mean squared error (RMSE), and mean absolute percentage error (MAPE), and it was found to be 0.019, 0.139, and 0.0983, respectively. This study concluded that ANFIS was able to model pavement conditions effectively.