Pneumatic non-contact topography characterization of finish-ground surfaces using multivariate projection methods

This paper reports on the application of multivariate analysis methods for the non-contact topography assessment of finish-ground surfaces of roughness in the range of 0.1–0.8 μm Ra. The roughness information is extracted from the frequency spectrum of the back pressure signal acquired using a pneumatic gauge as the surface traverses past the nozzle. Principal components analysis is demonstrated to be effective in the unsupervised classification of lapped and ground surfaces of an identical nominal roughness of 0.1 μm Ra, even under conditions that the corresponding frequency spectra are contaminated with noise and affected by vibration. Projection to latent structures analyses are further shown to be capable of discriminating cylindrical ground surfaces based on along-the-lay measurements from a rotating component, and formulating multivariate regression models appropriate for process monitoring.