نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشکده عمران، دانشگاه صنعتی امیرکبیر
2 دانشکده عمران امیرکبیر
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Linear and alligator cracking are critical indicators of asphalt pavement performance. The accurate prediction of these ciritical cracking distresses are of significant importance in effective and efficient pavement maintenance planning. This study proposes a data-driven framework based on multimodal data from the Long-Term Pavement Performance (LTPP) database, incorporating traffic, climatic, and performance-related variables to predict distress severity, length, and area. Key features, including surface distress indices, overlay thickness, traffic characteristics, and climatic indicators, were extracted and refined through feature engineering. Machine learning-based models were developed for severity classification and quantitative distress prediction using Artificial Neural Networks (ANNs). Addressing class imbalance with SMOTE improved severity classification accuracy from 0.782 to 0.843 for linear cracking and from 0.845 to 0.930 for alligator cracking. The models demonstrated strong predictive performance, achieving R² values of 0.941 for linear crack length and 0.954 for alligator crack area, supporting their applicability in preventive maintenance and pavement life-cycle management.
کلیدواژهها [English]
)