Comparison of mode decomposition methods tested on simulated surfaces

Abstract

Multiple modal decomposition of surfaces methods are increasingly used to analyse the typical geometric defects of manufactured surfaces. According to the context, this decomposition can either be done on a base which is known a priori (e.g. Discrete Cosine Transform, natural vibration modes etc.) or on a base that is identified from a set of measured surfaces (i.e. manufacturing dependent “technological modes” using the Principal Components Analysis or Independent components analysis). In this paper, a set of simulated surfaces are generated by linear combination of a given typical defect set in order to compare the efficiency the two different techniques. The compared techniques are, 1) methods founded on an a priori base and, 2) multivariate analysis methods (The key modes are identified for each method and compared to the technological modes used to generate the trial surfaces. From this study it may be concluded that while the first method does not allow the identification of the technological modes, the second does provide possible insight into the production technologies.