This paper presents a method of applying wavelets to decompose three-dimensional surface into multiple-scale subsurfaces, and of using the subsurface features to predict surface functions and detect machining errors. The one-dimensional discrete wavelet decomposition is first introduced, and then, it is extended to decompose and analyze three-dimensional surfaces. In this study, applications of wavelets decomposition are demonstrated in several automotive case studies, including abrupt tool breakage detection, chatter detection, cylinder head sealing/mating surface leak path detection, and transmission clutch piston suiface none lean up region detection. These case studies successfully demonstrate that the proposed multiple-scale two-channel wavelet decomposition method can be served as a useful tool for surface functions prediction and machining errors detection.

Optical measurement techniques are often employed to digitally capture three dimensional shapes of components. The digital data density output from these probes range from a few discrete points to exceeding millions of points in the point cloud. The point cloud taken as a whole represents a discretized measurement of the actual 3D shape of the surface of the component inspected to the measurement resolution of the sensor. Embedded within the measurement are the various features of the part that make up its overall shape. Part designers are often interested in the feature information since those relate directly to part function and to the analytical models used to develop the part design. Furthermore, tolerances are added to these dimensional features, making their extraction a requirement for the manufacturing quality plan of the product. The task of "extracting" these design features from the point cloud is a post processing task. Due to measurement repeatability and cycle time requirements often automated feature extraction from measurement data is required. The presence of non-ideal features such as high frequency optical noise and surface roughness can significantly complicate this feature extraction process. This research describes a robust process for extracting linear and arc segments from general 2D point clouds, to a prescribed tolerance. The feature extraction process generates the topology, specifically the number of linear and arc segments, and the geometry equations of the linear and arc segments automatically from the input 2D point clouds. This general feature extraction methodology has been employed as an integral part of the automated post processing algorithms of 3D data of fine features.

This research presents a new way to determine the condition of a cutting tool based on high definition surface texture parameters. Recently, a laser holographic interferometer has been developed to rapidly measure the whole workpiece surface (e.g. 300mm×300mm) and generate a 3D surface height map with micron level accuracy. This technique enables on-line surface measurement for machined parts. By measuring the surface texture of workpieces, the interaction between the tool’s cutting edges and the surface can be extracted as a spatial signature. It can then be used as a warning sign for tool change because the workpiece produced by a heavily worn tool exhibits more irregularities than those produced by a normal tool. Three surface texture parameters: image intensity histogram, surface peak-to-valley height and surface waviness are employed to detect the onset of severe tool wear. Furthermore, surface waviness can also be used to classify the different phases of tool wear. In this work, nine surface samples under different tool wear phases are created and analyzed using surface texture parameters combined with Statistical Process Control (SPC) charts to assess tool conditions. The results verify that these surface texture parameters can be used for on-line tool wear monitoring.

No abstract

No abstract