Advanced signal processing of wayside condition monitoring of railway wheelsets

Giannouli, Eleni (2021). Advanced signal processing of wayside condition monitoring of railway wheelsets. University of Birmingham. Ph.D.

Preview

Giannouli2021PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (9MB) | Preview

Abstract

Railway transport is an efficient and environmentally benign method of transport. With global warming effects intensifying it has become more urgent that mobility and economic prosperity are maintained by delivering increased transport efficient. Hence, railway transport has a significant role to play in the forthcoming decades.

Punctuality and safety of railway operations is critical in ensuring unhindered transportation for passengers and freight. Rolling stock are required to operate at higher speeds and carry heavier axle loads than ever before. This puts increased pressure to rolling stock operators and infrastructure managers in trying to avoid disruption and potential accidents which also leads to higher transportation costs.

Remote condition monitoring has increased in significance for railway transport over the last few years. However, there are still a lot to be done before breakthrough remote condition monitoring technologies are delivered at commercial scale in the wider international railway network.

Different remote condition monitoring systems are installed wayside in order to evaluate the structural integrity of rolling stock wheelsets, detect any potential rolling stock fault in time and minimize the likelihood of a serious railway accident. The existing wayside condition monitoring system are based on infrared cameras, acoustic arrays and strain gauges. Despite significant investments by the rail industry in this area, false alarms can still occur and many of condition monitoring systems are able to detect faults once they become critical.

In the present thesis, a novel approach based on integration of acoustic emission and vibration analysis together with advanced signal progressing is detailed. Tests ranging from laboratory tests under controlled conditions, all the way up to trials under actual operational conditions in the UK network have been carried out, yielding promising results. The experimental methodology employed has shown that acoustic emission is particularly efficient in detecting and ranking potential axle bearing defects. When acoustic emission is coupled with vibration analysis, it is possible to detect axle bearing defects whilst avoiding misinterpretation of wheel flats for axle bearing defects. The results obtained suggest that the widespread use of the reported methodology in the railway is feasible.

The novel RCM system can enhance the reliability, availability, maintainability and safety of rolling stock wheelsets. Experimental work have been carried out under actual operational conditions in UK rail network at Cropredy, at Chiltern Railway line. The novel RCM system has been installed adjacent to Hot Box Axle Detector for comparison purposes. No interference on the track circuits is the main advantage of the proposed system. During the signal processing module of the system, freight and passenger train waveforms were identified to contain evidence of potential bearing faults. The results still require follow up validation from Network Rail.

Time, frequency and time-frequency analysis have been applied to the acquired data. High amplitude peaks and signal modulation were visible at raw data. The acquired signals were transferred to frequency domain. Harmonics in frequency distribution were clearly seen. These frequency bands can be used as a reference for the band pass filter at HFRT process.

HFRT algorithm has been effectively applied in the captured data in order to identify the fundamental fault frequency and its harmonics. Sidebands were also visible. TSK analysis was also applied in the raw signals. Frequency bands with high kurtosis values can be used as a reference for further analysis.

In addition, laboratory experiments at University of Birmingham and Long Marston trials under controlled conditions have been carried out in order to evaluate the reliability of the system in early diagnosis of wheel and axle bearing defects. Acoustic emission and vibration signals have been collected. From the results obtained, it has effectively demonstrated that fault detection can be achieved using the frequency distribution of the signal. Defect type evaluation can be carried out by detecting the fundamental fault frequency at the HFRT process and fault quantification can achieved by Normalized Moving RMS analysis.

In summary, the research contribution of this work is presented below:
\(\bullet\) Development and assessment of the vibroacoustic condition monitoring system for railway wheelsets. Experimental methodology and results considered in this study are the main contributions in the literature of this field.
\(\bullet\) In service passenger and freight trains have been monitored. Detection of potential bearing faults has been achieved.
\(\bullet\) Novel methodology applied at the acquired in-service data in order to determine the appropriate frequency range for the band-pass filter during the HFRT process. Frequency bands with high kurtosis values can be used as a reference for the band-pass filter. In addition, harmonics have been presented in frequency distribution of the signal. The frequency bands that harmonics were appeared can also be used for the design of the band-pass filter.
\(\bullet\) Comparison between advanced signal processing techniques using laboratory, in-field and in in-service signals. Detection of wheelsets faults, identification of type of the defect and quantification of fault severity can be achieved when combination of algorithm is applied at raw signals.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):