Quantifying the damage of in-service rolling stock wheelsets using remote condition monitoring

Krusuansombat, Panukorn (2023). Quantifying the damage of in-service rolling stock wheelsets using remote condition monitoring. University of Birmingham. Ph.D.

Preview

Krusuansombat2023PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (10MB) | Preview

Abstract

The global railway network is set to continue to expand in terms of size, passenger numbers and freight tonnage in the coming decades. The occurrence of derailments can lead to major network disruption, significant financial losses, damage to infrastructure and rolling stock assets, environmental damage, and possibly fatalities and injuries. Defects in rolling stock wheelsets can potentially result in severe derailments if left to grow to a critical level. Rolling stock wheelsets are maintained using preventative maintenance techniques. Predictive maintenance solutions prevent unexpected failure, boost operational efficiency, and lower costs.

The railway industry has been looking into the development of advanced and effective condition monitoring with a low capital cost for the online and real-time assessment of the rolling stock wheels' structural integrity and subcomponents (wheels, bearings, brakes and suspension). Existing wayside measurement systems are based on different technologies, including hot boxes, acoustic arrays, wheel impact load detectors, etc. However, significant flaws, especially bearing failures, are challenging to identify. Hot boxes can only detect bad bearings after they overheat. This indicates that the bearing has failed and will be seized soon.

The combination of acoustic emission (AE) and vibration analysis has been used in this study to identify wheelset defects, particularly in wheels and axle bearings. Based on the new approach and thanks to the capability of early fault detection, predictive maintenance methods can be effectively applied whilst minimising the risk of catastrophic failure and reducing the level of disruption to an absolute minimum.

The present study looked into the quantitative evaluation of damage in axle bearings using an advanced customised vibroacoustic remote condition monitoring system developed at the University of Birmingham to improve the early fault detectability in in-service rolling stock wheelsets and improve maintenance planning. Laboratory tests using AE sensors and accelerometers were conducted to compare the sensitivity of each technique and evaluate the synergy in combining them. An experiment using the Amsler machine and bearing test rig proved that raw data and Fast Fourier transform (FFT) are inefficient for defect detection. More advanced signal processing techniques, including Kurtosis, were also applied to find the ideal core frequency and bandwidth for a band-pass filter. Cepstral analysis determines the complex natural logarithm of data's Fourier transform, and the power spectrum's inverse Fourier transform. It helps identify the bearing defect's harmonics from vibration measurement.

High-frequency harmonics arising from wheel and axle bearing faults were proven to be detectable from the acquired AE signals. The trial at Bescot yard demonstrates wayside measurement using a compact data acquisition system. Kurtogram-based band-pass filters eliminate environmental and undesired vibrations. The filtered signal with a better signal-to-noise ratio has less noise than the original signal. Another real-world wayside measurement was conducted at the Cropredy site to demonstrate train and wheelset defect detection.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):