Hybrid data fusion strategy for the low-uncertainty 3D calibration of cylinder standards

Abstract

A hybrid data fusion strategy which has been developed to fuse the form and diameter data acquired in the low-uncertainty calibration of cylinder standard is presented. It consists of two Gaussian process models and one least squares model. It is intrinsically robust and, as such, not sensitive to outliers and data randomness. The fused data points reconstruct the 3D cylinder form and the estimated parameters are used for profile transformation. The transformed profiles can be applied to other calculations. Simulations were conducted to test the data fusion performances. The results showed that the data fusion error was as low as 20% of the measurement uncertainty. The data fusion process largely reduced the uncertainty of data, namely, the uncertainty of the fused data was only 20% to that of the raw data. Experiments were performed by applying the data fusion strategy to the calibration of a piston gauge standard. The data fusion results showed good agreement with the specified tolerances, which indirectly verified the good quality of the measurement. The hybrid data fusion strategy is suitable for generalized calibration of cylinder standards and each step of it can also be applied independently to the data fusion of roundness profiles with diameters or with straightness profiles.

Simulation model for uncertainty estimation of measurements performed on five-axis measuring systems

“…Uncertainty estimation for measuring systems is still a challenging task [1][2][3][4]. Among the developed solutions, simulation methods show significant promise, since they are designed to shorten the length of the quality control process and enable automated evaluation of measurement uncertainty.…”