Systems and Methods for Real-Time Measurements of Rail Vehicle Wheel Flange Wear using Inductive Displacement Sensor Under Effects of Temperature and Noises
No Thumbnail Available
Date
2021-08
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Railway systems are expected to be in their most safe state ensuring safe operations of trains. The contact between rail and wheel is one of the most fundamental aspects of railway systems. Therefore, the instruments and measurement devices giving essential information on wheel flange wear are expected to perform accurately with high precision. This research presents an online wheel flange measurement system based on inductive sensing technology. Previous research on using the inductive displacement sensor for wheel flange wear measurement has focused on the light rail vehicles, and considered the effect of temperature negligible, as opposed to mainline vehicles where the wheel temperature variation is large. Also, the filtering of noises has been done using non-real-time data. Therefore, the objective is to study the effects of temperature on the inductive displacement sensor, employ methods to reduce measurement drifts from temperature changes, perform multibody system simulations on the locomotive, design an online filter, and design a sensor support structure mechanism to be fixed on the locomotive. Several experimental measurements are carried in the lab on a moving disk. During measurements, temperatures around the inductive displacement sensor and the metal target are increased from 19℃ to 98℃, similar to temperatures built up by flange and tread contact in the rail. Then multiple regression analysis of the data is carried out to come up with the measurement model equation. The system uses machine learning algorithms to automate regression models built from the input data that have been taken from the inductive sensor, thermocouple, and micrometer readings. Using the regression model algorithm, which is uploaded in LabView, the real-time non-linear data correlation takes place, giving a clearance response signal with improved error from 3.8174% to 3.46646%. Effects of vehicle dynamics and sensor noise are removed from the measurements by designing an infinite impulse digital filter. The filter specifications are obtained by carrying out Fast Fourier Transform of the results from a locomotive multibody dynamic simulation along a curved track and the measurements data from the prototype. Experimental results show effects of temperature on the inductive sensor measured data which are eliminated using a machine learning algorithm. This effect is taken into account to quantify the clearance between the disk and the sensor tip. The precision and accuracy after filtering are determined to be 0.06 Volts and 0.0285 mm, respectively. This system is expected to enhance the real-time and/or online monitoring of the safety of rail vehicles. Also, it can be integrated with Automatic Train Protection (ATP) and the detection of track lateral irregularities from the wheel flange real-time measurement data.
Description
Keywords
Effect of temperature, FFT, inductive displacement sensor, sensor fusion, data acquisition, machine learning, multibody system dynamic simulation, online filtering, rail vehicle safety, wheel flange wear