Evaluating the measurement capability of a wheel profile measurement system by using GR&R

Reliable data with less variation play a key role for acceptance of the usefulness of the measurement output of a wheel profile measurement system (WPMS) in a railway network. However, in practice, most studies are carried out without checking the reliability of data from such a system, which may lead to inappropriate maintenance strategies. To ensure the measurement capability of WPMS and to support robust maintenance in railway systems, this study has evaluated measurement data for the flange height, flange thickness, flange slope, and tread hollowing of rolling stock wheels by using gauge repeatability and reproducibility (GR&R). In this study, acceptance and rejection criteria for the precision-to-tolerance ratio (PTR), signal-to-noise ratio (SNR), and discrimination ratio (DR) have been employed to evaluate the measurement capabilities. For the purpose of illustration, we have implemented a new proposed approach. This approach involves both an analysis using graphs with four regions with a confidence interval (CI) of 95% and an analysis using a graph with three regions with only the predicted values; the latter type of graph represents an innovation made in this study. This graph has the advantages that it can visualize three different levels of data quality in same figure, namely “unacceptable”, “acceptable” and “good”, and also include a number of measures without becoming unclear, which are features that have been missing in previous presentations. The results show that the measurements of the flange slope are on an acceptable level, while those for the flange height, flange thickness and tread hollowing have to be rejected as unacceptable. The action proposed for increasing the quality of data on the flange height, flange thickness and tread hollowing is to enhance the calibration of the WPMS. In conclusion, GR&R is a useful tool to evaluate the measurement capability of WPMS and to provide helpful support for maintenance decision making. This investigation also shows that there is good reason to be careful when selecting measures and when interpreting the results, since, for a certain wheel profile parameter, when one measure is used, the results may be acceptable, but when another measure is used, the results may be unacceptable.