Effect of the number of projections on dimensional measurements with X-ray computed tomography

In X-ray computed tomography (CT) reducing the number of projections (N-p) acquired for data reconstruction will reduce the measurement acquisition time and, thereby, the cost of the measuring process. However, reducing Np can also reduce reconstruction quality and the accuracy of dimensional information provided by the CT measurement. This paper assesses changes in dimensional accuracy of X-ray CT data as a function of N-p. The variance of CT dimensional measurements, with respect to reference data obtained from tactile coordinate measurement machines (CMMs), is studied. Two cheese-like hole-cubes made of aluminum material and nylon (a polyamide thermoplastic) are used as measuring workpieces. It is determined that using N-p between 600 and 2000 does not produce major changes in accuracy for size measurements (lengths and diameters), for which absolute deviations between CT and reference data were mainly within the range of 5-10 mu m. This range reflects sub-voxel accuracies of CT measurement (the voxel size for CT data reconstruction was 73 mu m) when determining component size dimensions. However, for N-p < 600 the accuracy of CT measurements rapidly deteriorates, with deviations that can be much larger than 100 mu m. This is worse for measurements of form (flatness and cylindricity). Reducing N-p strongly affects form measurements. To assess form properties of CT reconstructed surfaces, using a large N-p (> 2000) is preferable. Although the accuracy of measurements of size also deteriorates with reducing N-p the accuracy loss is less severe, particularly if averaging over several data points. When measuring size, a loss of image quality can be tolerated as a trade-off for time optimization in CT data collection. Assessments of image quality further complement the conclusions presented in this paper.