Performance study of dimensionality reduction methods for metrology of nonrigid mechanical parts

Abstract

The geometric measurement of parts using a coordinate measuring machine (CMM) has been generally adapted to the advanced automotive and aerospace industries. However, for the geometric inspection of deformable free-form parts, special inspection fixtures, in combination with CMM’s and/or optical data acquisition devices (scanners), are used. As a result, the geometric inspection of flexible parts is a consuming process in terms of time and money. The general procedure to eliminate the use of inspection fixtures based on distance preserving nonlinear dimensionality reduction (NLDR) technique was developed in our previous works. We sought out geometric properties that are invariant to inelastic deformations. In this paper we will only present a systematic comparison of some well-known dimensionality reduction techniques in order to evaluate their accuracy and potential for non-rigid metrology. We will demonstrate that even though these techniques may provide acceptable results through artificial data on certain fields like pattern recognition and machine learning, this performance cannot be extended to all real engineering metrology problems where high accuracy is needed.