AI-Enabled Pavement Condition Modeling and Maintenance Prioritization for Sustainable Road Infrastructure: An Integrated Weibull–Fuzzy–Machine Learning Framework

Pavement-management systems rely on effective condition assessment and maintenance prioritization, yet existing frameworks rarely integrate the three complementary paradigms of probabilistic deterioration modeling, interpretable rule-based prioritization, and data-driven prediction within a single workflow. Moreover, studies that combine fuzzy prioritization with machine learning often reuse fuzzy-derived priority scores as predictive inputs, creating target leakage and potentially overstating predictive performance. To address these gaps, this study develops a novel integrated framework combining Weibull reliability analysis, Mamdani fuzzy inference, and ensemble machine learning while explicitly preventing target leakage. The framework was evaluated using 27,701 pavement observations collected from Road No. 65 in the Qassim Region, Saudi Arabia, between 2018 and 2022. Results show that Weibull analysis captured lane-specific condition differences, with shape parameters ranging from 3.55 to 6.79 and scale parameters from 81.19 to 93.06. A nine-rule Mamdani fuzzy system integrated Pavement Condition Rating (PCR) and International Roughness Index (IRI) to generate a continuous maintenance-priority score. Random Forest and Gradient Boosting models predicted PCR using only physically independent condition indices and outperformed a linear baseline. The novelty of the proposed framework lies in its integration of statistical reliability analysis, interpretable fuzzy decision logic, and leakage-free machine learning prediction within a unified pavement-management workflow. The framework supports systematic condition assessment and maintenance prioritization, while future work should incorporate economic, life-cycle, and environmental factors and validate predictive performance on additional road networks.