Comparison of Correlation Methods for Profile Measurements

摘要

This paper addresses the problem of quantifying the similarity between profile measurements that are acquired as sets of discrete data points. Several different time-domain correlation methods are presented. The methods are compared using profile measurements of the raceways of bearing rings obtained with a post-process measurement machine. The machine yields a continuous measurement of a part for several revolutions. Due to the machine design, the number of data points per revolution varies slightly. It is shown that the correlation methods are capable of determining the exact locations where the profile repeats itself. This is experimentally confirmed for different part geometries. Using a suitable interpolation algorithm, the individual measurements are interpolated to an equal number of data points which allows averaging them into a single measurement. It is demonstrated that the averaged measurements yield a lower measurement error than the individual measurements and are in better agreement with reference measurements.