Detection and Correction of Abnormal Data in On-Machine Measurement of Wheel Tread Profile

This study is a continuation of the findings of previous studies in the measurement data of wheel tread profile, abnormal data represent an important factor influencing the measurement result. To reduce the impact of abnormal data and improve measurement accuracy, this study focuses on the detection and correction of abnormal data in the measurement of the wheel tread profile. The detection, identification, correction, and confirmation of the abnormal data in the measurement data were systematically investigated. Regarding the detection of abnormal data, a nonstatistical algorithm with profile fitting was proposed to identify outliers in the measurement data. To correct abnormal data, correction strategies based on Kalman’s prediction have been proposed to identify and compensate for outliers. Synthetic simulations and experiments are performed to validate the proposed detection and correction methods. The results indicated that the abnormal data can be not only detected but also corrected well, and the data reconstruction deviation achieved a satisfactory level. Accurate detection and correction of abnormal data in the measurement of the wheel tread profile were achieved, which is significant for the morphology measurement of wheels and safety evaluation of track vehicles.