Open Access

Anthony Orth

• Industrial X-Ray computed tomography (XCT) is an alternative procedure to tactile coordinate measuring machines (CMM) for dimensional measurement as it has the advantage to be able to measure internal and external features. However, the uncertainty determination according to the standard VDI/VDE 2630 Part 2.1 is both complex and resource consuming. The procedure could be simplified with the help of computer simulations if a suitable digital replica of the XCT machine is available. The project CTSimU2 has the aim of establishing a guideline for evaluating the applicability of such a digital replica. In this context, the work presented here focuses on: determining the sensitivity of two specimens, on several modelling parameters; determining the significance of the modelled parameters; and estimating the sensitivity of the specimens measurands. The approach was to determine the effects of selected parameters and parameter interactions on dimensional measurands of the specimens byIndustrial X-Ray computed tomography (XCT) is an alternative procedure to tactile coordinate measuring machines (CMM) for dimensional measurement as it has the advantage to be able to measure internal and external features. However, the uncertainty determination according to the standard VDI/VDE 2630 Part 2.1 is both complex and resource consuming. The procedure could be simplified with the help of computer simulations if a suitable digital replica of the XCT machine is available. The project CTSimU2 has the aim of establishing a guideline for evaluating the applicability of such a digital replica. In this context, the work presented here focuses on: determining the sensitivity of two specimens, on several modelling parameters; determining the significance of the modelled parameters; and estimating the sensitivity of the specimens measurands. The approach was to determine the effects of selected parameters and parameter interactions on dimensional measurands of the specimens by simulation. These effects were determined by varying the parameters in accordance with a D-optimal experimental design, with 83 runs for 10 parameters. The measurands were distances, flatness of fitted planes, cylindricities of fitted cylinders, and testing of modular transfer function on edges. By definition the effects are linked to the coefficients of the statistical model – multiple linear regression – related to the design criterion. The results yielded good models with high regression coefficients. The models for distances were dominated by the geometrical parameters, whereas parallelism and flatness responded more to x-ray parameters. The results have shown that both specimens are necessary to test the digital replica.…