| Author: | Chen, Long |
| Title: | Localization and characterization of the fault in wheel/rail system |
| Advisors: | Choy, Yat-sze (ME) |
| Degree: | Ph.D. |
| Year: | 2020 |
| Subject: | Railroad cars -- Wheels -- Defects -- Testing Fault location (Engineering) Railroads -- Safety measures Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Mechanical Engineering |
| Pages: | xxv, 124 pages : color illustrations |
| Language: | English |
| Abstract: | This thesis presents a microphone array-based approach to localize and characterize the fault for the wheel/rail system under an environment with high background noise and the ground reflection interference. In the modern city nowadays, fault detection systems are typically applied in the railway industry to examine the structural health status of the wheel/rail system. The damage or fault on the surface structure of the wheels can cause impact noise lead to great environmental pollution or even cause fatal railway accident. Therefore, it is necessary to develop a structural condition monitoring system in order to improve the operability and reliability of the wheels in service. To deal with such problems, various methods have been employed to monitor the condition of wheel/rail system, such as acoustic emission (AE), ultrasonic wave, and magnetic testing. However, these methods can only indicate the existence of faults, but they cannot determine the position of the fault wheel. Since these existing techniques only rely on contact inspection methods and periodical maintenance to eliminate the potential security issues, this paper attempts to achieve fault visualization by adopting an acoustic-based noncontact diagnosis method. In this regard, the approach by using microphone array can be very effective to detect the faults on the structure surface. Before going into the main part of the thesis, preliminary investigation is first conducted by using a useful tool for characterizing and detecting non-stationary impulsive signals named spectral kurtosis (SK), then the wavelet transform (WT) based SK is proposed to extract the impulsiveness feature of signals in frequency-domain by smoother curves. However, considering the broadband feature of the impact noise generated by the uneven structural surface, it is difficult to identify the appropriate frequency bands for calculating the SK of a broadband signal. The research findings indicate that the wheel faults can hardly be identified using the traditional frequency-domain method. Thus, through the exploration of the traditional beamforming methods including delay-and-sum (DAS), linearly constrained minimum variance (LCMV) and multiple signal classification (MUSIC) by adopting plane wave and spherical wave model, a so called broadband weighted MUSIC (BW-MUSIC) method is proposed to detect the broadband signals. Meanwhile, a kurtosis based beamformer in time-domain is proposed to identify the fault under multiple sound sources circumstance and achieve fault visualisation as a function of time and space, instead of frequency and space. By using kurtosis beamformer, different types of faults on the wheels could be identified and localised by observing the kurtosis value on the beamforming sound map. The effectiveness of the proposed method to diagnose the wheel fault has been validated experimentally in this section. By the validation of experiments and simulations, the proposed kurtosis beamformer in time-domain is feasible to distinguish the location of the wheel with fault from the pristine wheel. The application of kurtosis enables the identification and localisation at extremely low signal-to-noise ratio (SNR). By thoroughly discussions and comparisons with other estimators such as peak value and root mean square (RMS), the kurtosis beamformer was found to be suitable for a wide range of SNRs, even in cases where the impulsive feature in the signal is completely indistinguishable using the traditional beamforming method from the background noise. Compared with the beamforming output evaluated by RMS and peak value, the kurtosis beamformer has a lower sidelobe level (SLL) and its ability to extract the impulsiveness emerging from the background noise is much stronger. Typically, the beamforming power can yield accurate positions of the sound sources in a field of multiple sound sources based on the acoustical power. Its suitability for a wide range of SNRs makes the main advantage when compared with the state-of-the-art structural condition monitoring non-destructive testing (NDT) techniques, apart from eliminating the necessity to attach the sensors to the structure. The influence of the duration of data acquisition on the performance of kurtosis beamformer has been discussed as well. Generally, when the duration of data acquisition for analysis contains at least one impulse signal, it would be accurate enough to identify the fault according to the kurtosis beamformer. However, if the number of impulses per second is high, the kurtosis value becomes low and may influence the accuracy of the fault localisation. The localization approach is then extended to sound wave propagation model with ground effect and moving source. With multiple sound sources and the interference of ground reflection, the mentioned BW-MUSIC method could provide a separate and distinct localization result in sound map to compute as an initial value. Based on that, the so-called Levenberg-Marquardt and Crank Nicolson (LM-CN) method can provide a preferable estimation result of ground impedance. Compared with the former LM approach to estimate ground impedance, the rate of convergence among all parameters is improved in LM-CN method. And the convergence values are closer to the actual values. Theoretically, in low Mach number cases, moving source problems could be solved by adopting an instantaneous frequency. Therefore, the BW-MUSIC method is first examined. A de-Dopplerization approach is adopted and combined with the kurtosis beamformer to deal with the high Mach number cases in time-domain. The experimental results presented in sound maps had a good matching with the actual conditions in terms of localizing and identifying. The localization results for three moving sources were acceptable, and the improved sound maps can identify wheel fault with the interference of background noise. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/10545