Methodologies for model parameterization of virtual CTs for measurement uncertainty estimation

Abstract

X-ray computed tomography (XCT) is a fast-growing technology for dimensional measurements in industrial applications. However, traceable and efficient methods to determine measurement uncertainties are not available. Guidelines like the VDI/VDE 2630 Part 2.1 suggest at least 20 repetitions of a specific measurement task, which is not feasible for industrial standards. Simulation-based approaches to determine task specific measurement uncertainties are promising, but require closely adjusted model parameters and an integration of error sources like geometrical deviations during a measurement. Unfortunately, the development of an automated process to parameterize and integrate geometrical deviations into XCT models is still an open issue. In this work, the whole processing chain of dimensional XCT measurements is taken into account with focus on the issues and requirements to determine suitable parameters of geometrical deviations. Starting off with baseline simulations of different XCT systems, two approaches are investigated to determine and integrate geometrical deviations of reference measurements. The first approach tries to iteratively estimate geometric deviation parameter values to match the characteristics of the missing error sources. The second approach estimates those values based on radiographs of a known calibrated reference object. In contrast to prior work both approaches only use a condensed set of parameters to map geometric deviations. In case of the iterative approach, some major issues regarding unhandled directional dependencies have been identified and discussed. Whereas the radiographic method resulted in task specific expanded measurements uncertainties below one micrometre even for bi-directional features, which is a step closer towards a true digital twin for uncertainty estimations in dimensional XCT.