Boyacioglu, Pelin (2018) Prediction of Rail Damage on Underground-Metro Lines. Doctoral thesis, University of Huddersfield.

Safety and reliability of rails primarily depend on detection, monitoring and maintenance of rolling contact fatigue (RCF) defects. Since when they are undetected and untreated, they can further propagate and increase the risk of rail failures. Thereby, infrastructure managers (IMs) tend to detect these cracks at an early stage in order to manage this risk. After detection, the growth of these cracks should be monitored and efficient maintenance should be carried out to prolong the rail life. However, this requires reliable and sufficient field data with accurate prediction models of RCF damage and its counterpart damage mechanism; wear. Although the current models, which are particularly used on real track conditions, focus on mainline routes and were often validated using rail surface observations, lesser emphasis has been placed on underground-metro system tracks and the use of non-destructive testing (NDT) measurements particularly the crack depth which is a key parameter in the assessment of crack severity and maintenance planning.

In recent years, London Underground (LUL) has put additional effort to improve their rail inspection practices to support the optimisation of their rail maintenance strategy. Besides the use of several different NDT techniques, the magnetic flux leakage based sensor is used to measure the depth of detected cracks. Research suggested that this rail inspection data could be used to improve the accuracy of damage predictions. With the help of successive measurements, the severity of damage could be quantified and the changes in RCF estimations and its interaction with wear over time could be demonstrated. It was proposed that this should increase the confidence in prediction models for maintenance planning and help to support future maintenance optimisation.

Owing to use of different NDT techniques, a significant volume of field defect data was collected and examined in detailed during the research to understand the dominant damage mechanisms and the influential factors promoting RCF crack growth. It was found that severe damage is not limited to mainline and freight routes, with rails on metro lines also suffering from a high number of cracks. In addition, the various track data consisting of wheel-rail profile measurements, track geometry, vehicle speed diagrams and traffic information were also submitted. This provided a good opportunity to build detailed vehicle dynamics simulations for the selected lines to be further studied on LUL.

The Whole Life Rail Model (WLRM) and the Shakedown Map were selected as predictive models since, they can integrate with vehicle dynamics simulations. However, when the main input of WLRM; ‘Tγ’ was initially applied to LUL tracks, it was found that while, it successfully showed the effect of significant factors, it resulted in over-and under-estimation of the RCF damage in several locations. Therefore, the research investigated the interaction between the model parameters and their comparison on sites with and without reported RCF defects to find an optimum solution. The results indicated certain distinctions and hence, new wear and RCF damage prediction methods were developed using a combined Shakedown Map and Tγ approach.

Both of the methods were applied to selected LUL RCF monitoring sites. Whilst the wear method was applied to predict the loss in cross-section area of the rail, the new RCF crack depth prediction model was validated using the MRX-RSCM crack depth measurements. The location and severity of both damage types were successfully predicted. To observe and predict the changes in RCF damage including the interaction of wear over successive measurement intervals brought novelty to the study. In addition, the accuracy of predictions was improved on sites with various track characteristics such as high and low rail of checked and unchecked curved track section and tangent tracks.

FINAL THESIS - Boyacioglu.pdf

Download (18MB) | Preview


Downloads per month over past year

Add to AnyAdd to TwitterAdd to FacebookAdd to LinkedinAdd to PinterestAdd to Email