Data-driven condition assessment of track substructure
Restricted (Penn State Only)
- Author:
- Nazari, Saharnaz
- Graduate Program:
- Civil Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- May 31, 2024
- Committee Members:
- Tong Qiu, Co-Chair & Dissertation Advisor
Hai Huang, Co-Chair & Dissertation Advisor
Ming Xiao, Major Field Member
Tieyuan Zhu, Outside Unit & Field Member
Farshad Rajabipour, Program Head/Chair - Keywords:
- SmartRock
Track substructure
track health monitoring
damage detection
Smartgrid - Abstract:
- The ballasted track is the most popular in North America and the most cost-effective, especially for running freight trains. In track, the substructure plays a critical role in maintaining track geometry and consequently affects the efficiency and safety of rail operations. The substructure is vulnerable to several defects, such as fouling, formation of pockets, poor drainage, and settlements. However, maintaining the substructure presents challenges, and timely detection of any adverse change in its condition is a significant area of focus and concern for the railroad industry. Visual methods, as the primary way of substructure inspection, cannot catch defects. Other Traditional procedures for track substructure evaluation are invasive and invariably result in disruptions of train operations and increased financial expenditures for the industry. As such, track maintenance is usually scheduled and carried out at fixed intervals. However, these maintenance actions may not always be timely or target the right area or issue without knowing the exact condition of the substructure and its components. With recent advances in technology and the development of data analysis tools, the railroad industry has adopted innovative inspection and monitoring techniques for the track infrastructure as complementary to traditional methods in order to overcome some limitations inherent in the old regimes and gain more accurate insights into the state of the track substructure. These new track monitoring techniques employ sensors to gather the information that helps railroaders detect defects in time and adopt appropriate maintenance measures reducing unexpected breakdowns in services and saving time and money for the industry. Today, a large variety of sensor-based systems are available, each generating vast amounts of data. A new challenge for the railroaders is choosing the right monitoring system that encompasses the best instrumentation for gathering and processing the information so as to inform them of the appropriate maintenance measures to be undertaken. The objective of this research study is to make use of the advances in big data analytics and build and evaluate a data-driven model that works with two emerging monitoring technologies, namely the SmartRock and the Smartgrid, to develop a system for monitoring the track substructure. The developed model together with these monitoring systems proved to be capable of damage identification in an under-operation track. This study is a stepping-stone toward a real-time data collection and integrated analysis system, allowing railroad companies to more accurately identify the condition of their ballast and the trackbed. This information enables them to proactively assign maintenance windows that ensure safe and efficient train operation with the least amount of train delay due to maintenance outages. The first stage of this research study focused on the ballast layer and developing a data-driven method based on pattern recognition techniques to identify fouling in track using the SmartRock sensor. To this end, SmartRock was applied in a field test to record the railroad ballast particle movement in the revenue service lines where mud spots existed. The result of this study shows that this system can accurately characterize the ballast condition. Later, a similar methodology was applied to the data collected from another field experiment to distinguish the differences in ballast particle performance due to under-tie pad utilization in track. The analysis is performed by coupling Autoregressive Moving Average (ARMA) as a sensitive feature with a supervised classification technique. The result showed that SmartRock could measure the performance of railroad ballast particles with high reliability. The second part of this dissertation delves deeper into the substructure to provide a system (monitoring device together with an analysis methodology) to characterize the performance of the ballast and the subgrade in track. To this end, a new monitoring technology composed of a sheet of geogrid instrumented with strain gauges was developed – The same will be referred to as Smartgrid in this dissertation. The Smartgrid can measure vertical stress and horizontal strain at the ballast-soil interface at four different spots in the track. These spots are under the middle of the tie, under the railseat, at the tie end, and in the crib under the railseat. This system helps railroaders get an insight into the performance of the ballast and the soil separately, contrary to conventional parameters, such as the track modulus, which measures the track substructure's stiffness as a whole. The Smartgrid was tested in the laboratory. The laboratory experiment helped develop a statistical analysis methodology that together with data from the Smartgrid could characterize the health conditions of ballast and subgrade individually. Further, this system was also tested in a field on revenue service lines that had mud spots in order to verify its efficacy. This study's result shows the Smartgrid method's potential in facilitating maintenance planning of railway tracks for a safer and more efficient transportation system.