On-line tool misalignment detection using an imaging method

Abstract

The Opticam spherical surface generator is developed at the Center for Optics Manufacturing (COM). It combines computer-numerical-control (CNC) technology with bound abrasive ring grinding geometry to create advantages over the conventional loose abrasive grinding systems in flexibility, productivity and accuracy. The Opticam SX utilizes a ring grinding geometry to generate spherical surfaces. By adjusting the tilt angle and the diameter of the ring cutter, the Opticam is capable of generating spherical surfaces with an essentially unlimited range of radius of curvature. This ring grinding geometry, however, requires a precise alignment between the cutter and the generated part. Residual cutter misalignments in the machine setup cause the finished surface to deviate from the design shape. Current tool misalignment detection techniques seriously limit the productivity of the Opticam system, and a new metrology system is introduced in this dissertation. The new system is geometrical ray trace based, and has some similarities to the moire deflectometry techniques. It uses a non-contact method to measure surface slope errors from images reflected off Opticam generated surfaces. The detected slope errors are used to determine the corresponding tool misalignments in the Opticam generator. The system setup is simple and not sensitive to vibrations. This system is compatible with the wet grinding environment. Furthermore, this new metrology system is capable of measuring surfaces with a wide range of radius of curvature from convex to concave. A prototype system was built based on this technique. It has been evaluated on-line in the Opticam SX. A surface wetting technique is used to allow detection on Opticam generated surfaces independent of the actual surface finish. The experimental results suggested that the on-line detection system is capable of detecting tool misalignment corresponding to a peak-to-valley surface figure error that is 1 mum or greater on ground surfaces. Better than 0.5 μm peak-to-valley surface error detection was achieved on specular surfaces. It was also found that machine dwell produces surface figure errors that are opposite to the errors produced by y-direction tool misalignment. Best surface figures are achieved when machine dwell errors are balance by residual tool misalignments in the Opticam machine.