Roundness Filtration by Using a Robust Regression Filter

In roundness measurement, filtration is a basic step before the evaluation of the roundness deviation parameters. Currently the widely used phase-correct 2RC and Gaussian filters are not robust against measurement outliers. This paper introduces a robust roundness filtration technology which is based on the robust Gaussian regression filter. Mathematical solution and its algorithms calculated in time and frequency domains are both presented. Both simulated and practical roundness profile data have been used to test the speed and accuracy of the algorithm. Results show that with the proposed robust algorithm, the outlier's influence on the roundness deviation parameters can be reduced significantly, while still keeping the similar transmission characteristics with the standard Gaussian filter very well. Also, the difference between the results of the two presented algorithms is very tiny and negligible.