The axle bearing of a railway vehicle supports a high load between an axle and a bogie frame and plays a role in smooth rotation of a wheelset. Currently, maintenance methods to prevent failure of axle bearings of railway vehicles are as follows. When...
The axle bearing of a railway vehicle supports a high load between an axle and a bogie frame and plays a role in smooth rotation of a wheelset. Currently, maintenance methods to prevent failure of axle bearings of railway vehicles are as follows. When a certain mileage is reached, there is a method of checking the condition with the naked eye and refueling the grease inside, and a maintenance method of replacing it with a new product when the complete disassembly maintenance cycle is reached. In addition, the axle temperature detection device installed on the outside of the track monitors the temperature of the axle during operation, and when the temperature rises above a certain level, an alarm is generated so that immediate accident prevention and condition-based maintenance can be performed.
In the past, many studies have been conducted on the condition-based maintenance of axle bearings accompanied by the introduction of new equipment and the reliability improvement of axle temperature detection devices. Since the main point of this is to introduce additional facilities to check the state of parts in real time, there is a lack of research on how to utilize the existing facilities, the axle temperature detection device, which is a ground detection device, as efficiently as possible.
In this study, a method for more accurately diagnosing the failure of an axle bearing by utilizing the currently used axle temperature detection device was studied.
First, preprocessing is performed to standardize the temperature data of the axle bearings collected by the axle temperature detection device into Z-scores. Then, an outlier score established based on past alarms and maintenance implementation history was established, and an algorithm generating an alarm was created. An appropriate cumulative cycle was set based on the maintenance details, and a critical point for the outlier score was set. As a result, it was evaluated that it could generate a more accurate alarm than the actual maintenance history.