Over the past decade, urban transportation infrastructure has experienced rapid expansion. However, due to variations in deployment time and environmental conditions, noticeable discrepancies have emerged in both the operational status and data quality of traffic sensors, which severely hinder the deployment, development, and in-depth exploitation of traffic data. To address these challenges, this study proposes an evaluation and optimization framework for multi-source urban traffic sensing data. By integrating data-cleansing algorithms, the framework enables simultaneous data quality analysis and operational status assessment of heterogeneous sensors. A one-year empirical study was conducted using flow data collected from 91 detector stations across the urban road network. The framework successfully evaluated the operational status of all sensors and achieved a recovery accuracy of 97.47% under scenarios with limited data loss, while maintaining a stable accuracy of approximately 90% even in cases of severe data missing. The proposed approach provides a solid foundation for high-quality operation, maintenance, and data-driven applications of large-scale urban traffic sensor networks.