Safety IDEA Project 41 [Active (IDEA)]
Vibration-based Longitudinal Rail Stress Estimation Exploiting Contactless Measurement and Machine Learning
| Project Data
||University of Illinois at Urbana-Champaign|
||John S. Popovics|
In this study, the project team aims to develop new technology for in-place rail stress (rail neutral temperature or RNT) measurement that combines contactless acoustic sensing and machine learning technology. Safety is a principal concern of the railway industry, and track alignment irregularities caused by excessive rail stress can disrupt safe railway operation. Acoustic measurements of rail vibration provide a rich data set that contains information about the temperature and stress state of the rail, but also contains disruptions from other influences. The belief is that a machine learning approach can extract useful information about rail stress from the complex acoustic data set. Such a system could predict in situ rail stress or RNT in real time and without baseline measurements. Furthermore, fully contactless sensing enables the possibility of continuous inspection of large distances of track if deployed on a moving platform. The proposed work plan comprises two phases: 1) collect acoustic data using contactless sensing from locations of known in-service rail stress to build a massive training data set, and 2) design and implement machine learning protocols to accurately determine RNT from acoustic measurement data based on the trained model. If successful, the project will provide a practical and effective solution to a problem that continues to negatively affect railway operations because current technology does not support needed performance.