Development of an Adaptive Neuro-fuzzy Inference System (ANFIS) for Predicting Pavement Deterioration

Pavement maintenance is a critical aspect of transportation and infrastructure management, as it directly impacts traffic flow, vehicle maintenance, safety and accident rate. Effective prediction and prevention of pavement deterioration are essential for optimizing pavement maintenance strategies, reducing cost, and ensuring the lifespan or longevity of transportation. This study presents the development of Adaptive Neuro-Fuzzy inference system (ANFIS) for predicting pavement deterioration. The data used for this analysis is a historical data and field investigation data from the Cross River State pavement Maintenance Agency, Calabar, Nigeria. The ANFIS model was trained using a dataset with input variables, the outcome from the membership functions shows that in1=4.4, in2=3.15, in3=13.5 and output variable (out1=0.541). The model consists of 270 nodes, 33 fuzzy set, and utilizes sub clustering (FIS type), hybrid optimization method, Gaussmf membership function, and wtaver defuzzification. The study result showed that the ANFIS model accurately predicted pavement deterioration with Root mean Squared Error (RMSE) 0f 0.851716. The developed ANFIS model can be used by transportation agencies to predict pavement deterioration and prioritize maintenance activities. The Model’s ability to handle nonlinear relationships and uncertainty makes it a valuable tool for pavement management.