Automatic tolerance inspection through Reverse Engineering: a segmentation technique for plastic injection moulded parts

This work studies segmentations procedures to recognise features in a Reverse Engineering (RE) application that is oriented to computer-aided tolerance inspection of injection moulding die set-up, necessary to manufacture electromechanical components. It will discuss all steps of the procedures, from the initial acquisition to the final measure data management, but specific original developments will be focused on the RE post-processing method, that should solve the problem related to the automation of the surface recognition and then of the inspection process. As it will be explained in the first two Chapters, automation of the inspection process pertains, eminently, to feature recognition after the segmentation process. This work presents a voxel-based approach with the aim of reducing the computation efforts related to tessellation and curvature analysis, with or without filtering. In fact, a voxel structure approximates the shape through parallelepipeds that include small sub-set of points. In this sense, it represents a filter, since the number of voxels is less than the total number of points, but also a local approximation of the surface, if proper fitting models are applied. Through sensitivity analysis and industrial applications, limits and perspectives of the proposed algorithms are discussed and validated in terms of accuracy and save of time. Validation case-studies are taken from real applications made in ABB Sace S.p.A., that promoted this research. Plastic injection moulding of electromechanical components has a time-consuming die set-up. It is due to the necessity of providing dies with many cavities, which during the cooling phase may present different stamping conditions, thus defects that include lengths outside their dimensional tolerance, and geometrical errors. To increase the industrial efficiency, the automation of the inspection is not only due to the automatic recognition of features but also to a computer-aided inspection protocol (path planning and inspection data management). For this reason, also these steps will be faced, as the natural framework of the thesis research activity. The work structure concerns with six chapters. In Chapter 1, an introduction to the whole procedure is presented, focusing on reasons and utilities of the application of RE techniques in industrial engineering. Chapter 2 analyses acquisition issues and methods that are related to our application, describing: (a) selected hardware; (b) adopted strategy related to the cloud of point acquisition. In Chapter 3, the proposed RE post-processing is described together with a state of art about data segmentation and surface reconstruction. Chapter 4 discusses the proposed algorithms through sensitivity studies concerning thresholds and parameters utilised in segmentation phase and surface reconstruction. Chapter 5 explains briefly the inspection workflow, PDM requirements and solution, together with a preliminary assessing of measures and their reliability. These three chapters (3, 4 and 5) report final sections, called “Discussion”, in which specific considerations are given. Finally, Chapter 6 gives examples of the proposed segmentation technique in the framework of the industrial applications, through specific case studies.