Electronic Theses and Dissertations for Graduate School
Add My Work
Author Last Name
IDENTIFYING BALLAST FOULING USING STATISTICAL PATTERN RECOGNITION TECHNIQUES ON SMARTROCK DATA
Restricted (Penn State Only)
Master of Science
Date of Defense:
April 26, 2018
Dr. Hai Huang, Thesis Advisor
Dr. Tong Qiu, Thesis Advisor
Shihui Shen, Committee Member
Xinli Wu, Committee Member
Statistical Pattern Recognition Analysis
Railroad ballast serves different functions including draining water from track and distribution of the train loads. The ballast layer deteriorates and becomes fouled with time due to ballast particle abrasion and breakage as well as subgrade soil intrusion. Ballast fouling has become one of the most commonly seen track defects that can lead to inconsistent track performance. In the case of fouling, the ballast strength will decrease when it is wet (usually referred to as “mud-spot”) due to the lack of particle interlocking and lubrication effect of fine materials. However, both of the ballast strength and stiffness will increase dramatically when it is in dry condition as the ballast particles are well confined (Qian, 2016). This inconsistency in track behavior can cause higher deterioration rate of other track components such as rail and sleeper. Therefore, identifying mud spots in a timely manner is a critical issue in ballasted track maintenance. The main purpose of this thesis is using advanced sensor networks and statistical pattern recognition techniques to identify ballast fouling by studying the relationship between ballast fouling condition and ballast particle movement. To that end, several field experiments were carried out with the aim of monitoring and recording the particle movements under different ballast conditions. In particular, two sections with the same traffic but different track conditions: one with clean ballast and the other with mud pumping, were chosen. The SmartRock (Liu, 2015) is used to obtain ballast particle movement information under traffic. The SmartRock is a wireless sensor device built using the 3D printing technology and resembles the real ballast particles in terms of shape, inter particle friction and specific gravity. This sensor device has the ability to record translational and rotational movement of a single ballast particle under dynamic loading and transfer the real-time data via Bluetooth to a base station. The autoregressive (AR) model was then applied to each of the acceleration and rotation time histories collected from the SmartRocks embedded in the two sections, during which the autoregressive coefficients will be obtained. Those coefficients will serve as damage indicators to identify ballast fouling severities. The results and important findings are highlighted in this thesis.
Login using your Penn State access account to view the paper.