AI-POWERED RAIL TRACK AND ROAD POTHOLE FAULT DETECTION SYSTEM USING ADVANCED DEEP LEARNING TECHNIQUES FOR ENHANCED INFRASTRUCTURE SAFETY

Science and Tech

This study presents an AI-powered fault detection system for rail tracks and road potholes using a Modified Deep Convolutional Neural Network (DCNN) with Synthetic Minority Over-sampling Technique (SMOTE). The proposed model addresses critical challenges in infrastructure monitoring: detecting subtle defects and handling class imbalance in imbalanced datasets. By incorporating residual connections and attention mechanisms, the DCNN achieves superior feature extraction, while SMOTE significantly improves detection of minority fault classes. Experimental results demonstrate exceptional performance, with 98.2% accuracy for rail track faults and 97.5% for road potholes. The system achieves 96.5% recall for rail defects and 95.8% for potholes - a 15% improvement over baseline methods. With real-time processing at 45ms per image, the solution is deployable on edge devices for continuous monitoring. Key innovations include: (1) a novel DCNN architecture optimized for infrastructure defects, (2) effective SMOTE integration for class imbalance mitigation, and (3) comprehensive validation on diverse datasets. The system's high precision (97.8% for rails, 96.3% for roads) minimizes false alarms, while its recall ensures critical faults are rarely missed. This research contributes to safer, more efficient infrastructure maintenance by providing: (1) a robust AI framework for defect detection, (2) practical solutions for real-world deployment challenges, and (3) benchmarks for future work in smart infrastructure monitoring. The results highlight the potential of deep learning to transform traditional inspection paradigms, reducing costs while improving reliability.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.