Abstract
Additive manufacturing (AM) plays a vital role in realising the customised intelligent industry in Industry 4.0. However, quality control of AM products is one of the significant challenges and requires reliable dimensional metrology techniques to establish the link between part geometry, part quality and the build process. While conventional tactile and optical dimensional metrology techniques are limited by the complex free-form nature of AM parts, X-ray computed tomography (XCT) is considered a promising metrology tool for AM parts. XCT enables the non-destructive acquisition of three-dimensional holistic geometric information of an object, including external and internal structures, and has the potential to simultaneously perform different metrology tasks such as dimensional size measurement, surface texture measurement and porosity measurement. However, the application of XCT in the dimensional metrology of AM parts is restricted by its limited resolution, especially in the measurement of small-scale structures of dense metal parts, e.g., surface texture and pore structures. A thorough understanding is required for the evaluation of XCT resolution, the influencing factors of XCT resolution, the impact of XCT resolution on the measurement results and the improvement methods of XCT resolution. This thesis aims to investigate and improve the metrological structural resolution (MSR) of XCT to understand and enhance the performance of XCT on surface texture measurement of AM parts.Surface Amplitude Transfer Function (SATF) is proposed to evaluate the MSR of XCT. SATF is defined as the amplitude ratio of the spatial frequencies of the XCT measured surface to the spatial frequencies of the true surface, which characterises the magnitude part of the two-dimensional (2D) frequency response of the XCT measurement system. The calculation of SATF is based on the stochastic-deterministic AM surfaces. The calculation (denoising), validation and interpretation of SATF are presented. The wavelengths corresponding to the 80% SATF value are extracted as the evaluation metrics of the MSR of XCT.
The investigation of the influencing factors of the MSR of XCT is presented. The impacts of the measurement setups, including focal spot size (or current), voxel size (or magnification), linearisation-based beam hardening correction and reconstruction resolution, were investigated using SATF. An in-depth analysis of the impact of the measurement setups on the MSR of XCT is presented, and guidance on optimising the measurement configuration to improve MSR is provided. The results show that the MSR of XCT is linearly related to the focal spot size and voxel size, and the voxelisation effect causes a greater amplitude reduction effect on smoother surfaces. It suggests that the two factors need to be especially concerned when trying to optimise the MSR in the measurement configuration. The linearisation-based beam hardening correction is shown to significantly improve the MSR of XCT and is therefore recommended to be configured in XCT measurements. Using a reconstruction resolution higher than the voxel size is shown to slightly improve the MSR of XCT.
Further investigation of the influencing factors of the focal spot is presented, and guidance on minimising the focal spot in the measurement configuration is provided. The 2D effective focal spot (i.e., the point spread function, PSF) was reconstructed using the XCT projection image (radiograph) of a star pattern sample, and the influencing factors of the focal spot, including current, voltage, beam focusing and the use of physical prefilter, was investigated. The results show that current is linearly related to the focal spot size, and high voltage can significantly increase the focal spot size. Based on the relationship between current, voltage and focal spot size, a measurement routine is given for adjusting the current and voltage values to minimise the focal spot. Beam focusing is also shown to significantly affect the focal spot, therefore it is recommended to emphasise the filament alignment process for the beam focusing. The use of physical pre-filter is shown to have an insignificant influence on the focal spot. In addition, high current, high voltage and cold scan are shown to increase the instability of the focal spot, but the focal spot instability can be considered insignificant as the standard deviation of the focal spot size during a one-hour period is less than 1.5 m in this work.
Investigation of the removal of the focal spot blur from the XCT projection images to further improve the MSR of XCT is presented. The focal spot blurring effect on the XCT projection images can be mathematically modelled as the convolution of the true projection image and the 2D PSF, therefore, deconvolution can be applied to the projection images to reverse this process and restore the image information, leading to the restoration of the surface information and the improvement of the MSR. Investigations on the image restoration performance, the surface restoration performance and the performance optimisation of the projection-based deconvolution method are presented. It shows that the projection-based deconvolution method can restore surface structures that have even been lost in the original measurement and significantly improve the MSR of XCT. It also shows that severe image noise, large image sampling distance and significant image blurring limit the image and surface restoration.
In conclusion, this work provides an SATF method to evaluate the 2D MSR of XCT and presents two approaches to improve the MSR of XCT. One approach is to optimise the measurement configuration to minimise the focal spot. The other is to use deconvolution to reduce the focal spot blur on the projection images. The investigation of the MSR of XCT provides in-depth insight into the XCT measurement capability for AM surface texture, and the MSR improvement methods extend the measurement capability of XCT.
| Date of Award | 19 Mar 2024 |
|---|---|
| Original language | English |
| Sponsors | National Physical Laboratory |
| Supervisor | Jane Jiang (Main Supervisor), Wenhan Zeng (Co-Supervisor), Shan Lou (Co-Supervisor) & Paul Scott (Co-Supervisor) |