Generalised spatially weighted autocorrelation approach for monitoring and diagnosing faults in 3D topographic surfaces

Digital transformation driven by artificial intelligence (AI) allows Industry 4.0 and the internet of things (IIoT) to make significant advancements in automating, controlling, and improving the quality of numerous manufacturing processes. Three-dimensional (3D) surface topography of manufactured products holds important information about the quality of manufacturing processes. Surface topography consists of unique properties, which makes the current monitoring approaches ineffective in identifying local and spatial surface faults. In this paper, we develop a generalised spatially weighted autocorrelation approach based on AI for monitoring changes in products based on their 3D topographic surfaces. We propose two effective algorithms to identify and assign spatial weights to the topographic regions with suspicious characteristics. The normal surface hard thresholding algorithm initially enhances the representation of surface characteristics through binarization, followed by the normal surface connected-component labelling algorithm, which utilises the obtained binary results to identify and assign spatial weights to the suspicious regions. We then introduce a generalised spatially weighted Moran index, which exploits the assigned weights to locally characterise and monitor changes in the spatial autocorrelation structure of identified regions. After an anomaly surface is detected, we extract different fault diagnostic information. The proposed approach proves its robustness and efficiency in characterising, monitoring, and diagnosing different patterns of faults in 3D topographic surfaces.