Tramway infrastructure plays a key role in transporting large numbers of people in various cities worldwide. However, ensuring the quality and operation of the tram system is not easy because the tramway infrastructure is sensitive to degradation and faces different behaviours at different locations. For instance, transition zones are areas where local stiffness changes can cause a high track geometry degradation due to traffic loads and track settlement. To be well-informed about the health conditions of the infrastructure, structural health monitoring and maintenance are crucial for the inframanager. The literature has proposed that standard deviations of track gauge can indicate high geometry deterioration. However, geometry measurements are typically not conducted frequently, opening a gap in the lack of information on critical locations. In the field of conventional railways, using InSAR satellite data has been reported to be able to provide frequent data and to detect high settlement rates leading to high rates of track degradation. In this thesis, inspired by the experiences in the railways, an evaluation of InSAR satellite data will be conducted to perform a feasibility analysis for the case of transition zones in tramways.
This thesis focuses on predicting high settlement rates at transition zones using InSAR satellite data. Because InSAR data shows more frequent deformation data, it can potentially detect earlier high settlement rates or allow a more frequent data update of critical locations without requiring dedicated measurement trains. While geometry is the core in this thesis, InSAR could also be used for other applications such as vegetation control. In the case study, validation data with structural analysis and alternative means of condition measurement were unavailable. Thus, the development of an unsupervised approach that allows a first analysis to facilitate the exploration of the characteristics of the data and its relation with infrastructure has been considered.
Three transition zones of the tramway track in Amsterdam have been selected as case studies. The InSAR data points that are closer to the transition zone are extracted to use prediction methods. The coordinate position of these data points faces uncertainties. Therefore, different regions have been considered, and a method to obtain an area using the Pearson Correlation Coefficient from InSAR and GVB geometry data has been proposed. Then, two methods to predict high settlement rates have been proposed. The first method is the stochastic rates method, which uses half-year InSAR data settlement rates to indicate if a rate is higher than the historical rates. The other method uses data from different years to show a pattern and looks for dips to indicate these as high settlement rates. Finally, the method proposes using different data sources to enhance understanding of health conditions in transition zones. Speed plots were used to showcase the possibility of data fusion and correlation analysis, as driver behaviours at transition zones could be an interesting parameter to analyse.
In the results, the measured standard deviation of the track gauge of data points on the bridge was usually lower than the standard deviation of the track gauge of data points not on the bridge. Almost the same results were returned using the same principle with twist data. The uncertainty of both approaches has an order of 10-2, so they have similar performance. The stochastic rates method gives more precise and appropriate results than the prediction method based on BLUE. The two approaches have also been tested with a standard transition length of 10 meters at the Piet Wiedijkstraat and 20 meters for the other locations, instead of using the Pearson Correlation Coefficient method. The settlement rates on the bridge were mainly lower than those off the bridge. Also, the standard deviations of the track gauge on the bridge were lower than the standard deviations off the bridge. Subsequently, the stochastic rates method showed higher peaks than when using variable ranges. However, the prediction method based on BLUE still did not give a specific value like the stochastic rates approach. Finally, different data sources can be used to enhance the understanding of the behaviour of the transition zones. Speed plots have been used to reflect driver behaviour at transition zones. Still, other datasets like maintenance, bridge design, wear, precipitation, temperature, and vegetation could also be used in future research.