ADVANCED PREDICTIVE MODELING OF PAVEMENT CONDITION INDEX USING ARTIFICIAL NEURAL NETWORKS: A COMPREHENSIVE CASE STUDY IN TEXAS

Abstract

Roadways are pivotal to a nation’s economic growth, making effective pavement management systems (PMS) essential for maintaining highway networks cost-efficiently. This research explores the potential of forecasting the Pavement Condition Index (PCI) for flexible pavements using Artificial Neural Networks (ANNs) alongside the Long-Term Pavement Performance (LTPP) database. The study considers various factors, including age, traffic volumes, temperature, humidity, precipitation, freezing, and layer thickness (A.C. Base-Subbase-Subgrade). Utilizing data from 60 sections in Texas, USA, without overlays (classified as SPS-1 by LTPP), the ANN model was developed in MATLAB to predict PCI. The findings reveal the model’s exceptional accuracy in forecasting PCI, with correlation coefficients (R²) exceeding 90% for all sections. This indicates the robust applicability of ANN models in predicting pavement conditions, facilitating effective road maintenance planning.