Condition monitoring of curve squeal based on analysis of acoustic and vibration data

The railway industry is currently investing in condition monitoring techniques to be able to compete with other transportation mediums. One of the reasons for this investment is to be able to identify the incipient development of curve squeal in railway systems. The annoying high-pitched tonal noise produced because of curve squeal has necessitated the need for mitigation measures to be taken by railway operators. However, noise from the surroundings and other trains has affected the conventional use of microphones for monitoring curve squeal in tight curves. It is imperative that the railway industry introduce additional sensors to help in the characterization and identification of curve squeal in railway track as the train negotiates the curve.

The objective of this research is focused on the evaluation of condition monitoring performances using vibrations obtained from the wheel/rail roller and sound obtained remotely close to the wheel-rail interface to identify and characterize curve squeal. By the completion of the comparative studies, this research has resulted in a number of new findings that illustrate the significant contributions to knowledge. This research presents the application of correlation method to establish a reliable relationship between acoustic and sound for the detection and characterization of curve squeal on the twin disc rig. The sensors used to detect and characterize curve squeal are microphone and two accelerometers installed laterally on the wheel and rail roller rims. The contact conditions taken into consideration are dry contact, wet contact and friction modifier contacts. A MATLAB model was developed to detect and characterize curve squeal. The results of the simulated model showed some disparities between the simulated transition yaw angles and measured transition yaw angles for which curve squeal occurs. Time and frequency domain were employed to extract the features from the sensors. Correlation method was employed to classify the features extracted from the microphone and accelerometer data. The results obtained showed that a negligible or weak correlation coefficient value indicates the development of curve squeal on the twin disc rig in dry contact conditions. A moderate or strong correlation coefficient values is an indication of no curve squeal occur or curve squeal mitigation when contaminants (water and friction modifiers are introduced to the wheel-rail interface). The performance of the Correlation method for determining and classifying fault feature (curve squeal) extracted from the microphone and wheel/rail accelerometers has presented some useful qualities that makes it suitable in a real condition monitoring application system.