User-Driven Computer -Assisted Reverse Engineering of Editable CAD Assembly Model

This paper introduces a novel reverse engineering techniquefor the reconstruction of editable CAD models of mechanicalparts' assemblies. The input is a point cloud of a mechanicalparts' assembly that has been acquired as a whole, i.e. withoutdisassembling it prior to its digitization. The proposedframework allows for the reconstruction of the parametricCAD assembly model through a multi-step reconstructionand fitting approach. It is modular and it supports variousexploitation scenarios depending on the available data andstarting point. It also handles incomplete datasets. The reconstructionprocess starts from roughly sketched and parameterizedgeometries (i.e 2D sketches, 3D parts or assemblies)that are then used as input of a simulated annealingbasedfitting algorithm, which minimizes the deviation betweenthe point cloud and the reconstructed geometries. Thecoherence of the CAD models is maintained by a CAD modelerthat performs the updates and satisfies the geometricconstraints as the fitting process goes on. The optimizationprocess leverages a two-level filtering technique able to cap-ture and manage the boundaries of the geometries insidethe overall point cloud in order to allow for local fittingand interfaces detection. It is a user-driven approach wherethe user decides what are the most suitable steps and sequenceto operate. It has been tested and validated on bothreal scanned point clouds and as-scanned virtually generatedpoint clouds incorporating several artifacts that wouldappear with real acquisition devices