Automated assessment of measurement performance in optical coordinate metrology

The measurement of parts of high geometrical complexity, such as those fabricated via additive manufacturing technologies, represents a fundamental challenge for the quality assurance of manufactured components. The pressing need shared nowadays by numerous industrial sectors for obtaining increased product quality while reducing development times and costs, as well as achieving faster speed of production and inspection planning in mass customisation (ideally performed in real-time), are leading towards the design of smart measurement systems that, integrated directly in the production line, can achieve part quality inspection in a fully automated way. The future possibility of developing such “intelligent” instruments implies their capability of autonomously planning a measurement process and assessing measurement performance while the inspection task is in progress, making use of available pre-existing knowledge of parts, instruments and technologies, and employing smart algorithms for the optimisation of measuring procedures. In this context, optical coordinate measurement technologies appear as suitable candidates, due to their potential in featuring high densities in point-based sampling acquired at relatively fast rates, and in accessing complex surfaces despite line-of-sight issues. However, without metrics for quality, the employment and integration of such smart instruments cannot be fully accomplished. This thesis addresses the issue of quality in measurement, proposing algorithmic solutions to compute indicators of measurement performance directly from the measured point clouds and in a fully automated way. Starting solely from the knowledge acquired of the measured data and the underlying nominal geometry, these indicators are based on algorithmic point cloud processing pipelines, and make use of computational geometry and spatial statistics to primarily extract information about the quality of the measurement result. A first set of measurement performance indicators investigates the relationships between the measured point cloud and the reference geometry (in the form of triangle meshes) to automatically assess coverage and sampling density in relation to the individual surfaces of the measured part. Additionally, local dispersion of the measured points with respect to the underlying part region is evaluated. A second set of indicators investigates local dispersion of the point cloud, as well as local bias, by using a statistical point cloud models fitted to repeated measurement data. The second set of indicators is useful to assess metrological performance in repeatability or reproducibility conditions. The proposed sets of indicators are illustrated and validated through application to selected test cases of industrial relevance, generated via additive manufacturing technologies. The solutions developed and discussed in this thesis represent novel measurement performance assessment tools, which can be integrated into smart measurement systems. In the future, such instruments will be capable of self-assessing their own performance in-process (i.e., while measuring), and will be capable of planning the most suitable corrective actions, in the case that issues are detected in the quality of the measurement result (for instance, insufficient degree of coverage, unacceptable measurement error). Furthermore, such intelligent measuring systems will be suitable for integration with manufacturing machines, leading to the realisation of more flexible and more autonomous production systems