Comparison of nominal and real 2D contours of manufactured products using Ant Colony Optimisation of shape landmarks

Abstract

In studying manufacturing processes it is often necessary to compare nominal or reference product shapes that are expected to be ideally achieved after applying a nominal set of process parameters and the actual product shape achieved. The latter is usually obtained by metrological equipment such as profilometers, tomographs and 3D scanners, whilst the former may also result from manufacturing process numerical simulation software. In this work, comparison of 2D contours is studied by adopting the well-established Procrustes method. The method relies on minimizing deviation of landmark points belonging to the shapes under scrutiny. Landmarks that need to be compared pairwise should be chosen in an optimal way, especially when the contours contain different numbers of points, when they differ locally etc. Optimal determination of landmarks and as a result implementation of contour comparison is performed through ant colony optimization implemented on Matlab^TM platform. Indicative results are shown where deviations discovered signify shrinkage of a die-cast clay part.