Probabilistic Statistics-Based Endurance Life Prediction of Bridge Structures

Area of Science:

• Civil Engineering
• Data Science
• Structural Health Monitoring

Background:

• Bridge health monitoring traditionally focuses on damage detection and safety warnings using valid data.
• Evaluating fatigue in steel bridge panels is challenging due to intrinsic and extrinsic uncertainties.
• Existing systems often lack robust data processing and analysis for complex fatigue assessments.

Purpose of the Study:

• To develop a scalable and reliable big data platform for bridge health monitoring.
• To enable real-time data processing, cleaning, and analysis for enhanced safety warnings.
• To improve fatigue life prediction for bridge structures using probabilistic methods.

Main Methods:

• Construction of a big data platform using HDFS for storage and Spark for analysis.
• Implementation of Kafka for real-time data caching and Spark-streaming for processing.
• Development of data cleaning and missing data repair algorithms tailored to different data types (e.g., temperature, strain).

Main Results:

• The big data platform demonstrated high performance in offline and real-time processing, scalability, and fault tolerance.
• Optimal data cleaning methods were identified, addressing noise, jump points, and drift.
• Data patching algorithms achieved over 90% recovery accuracy for missing data.

Conclusions:

• The developed big data platform significantly enhances bridge health monitoring capabilities.
• Probabilistic fracture mechanics and big data statistics offer a superior approach to fatigue assessment compared to deterministic methods.
• The platform provides a reliable foundation for predicting bridge endurance life by integrating multiple influencing factors.