Improvement of surface form accuracy through iterative corrections


Student thesis: Master's Thesis

View graph of relations


  • Ka Lok YIU


Awarding Institution
Award date2 Oct 2008


Surface milling and grinding are widely used in the fabrication of free-form surfaces for molds and dies. The continuous demand for better product quality has led to demand in higher surface accuracy. Limiting factors are the positioning accuracy of the tool and the accuracy of the manufacturing process. Computer controlled surfacing (CCS) has been developed mainly for the fabrication of aspherical and freeform surfaces. The surface is repeatedly measured and corrected using abrasives until the target accuracy is attained. In principle, sub-micron accuracy is achievable. The accuracy is limited by surface measurement and not by the processing equipment. The current research investigates the adoption of CCP for the fabrication of freeform surfaces for molds and dies. The test-bed is based on a 6-axis RX robot as the motion platform with a spindle attached to the wrist of the robot. The positioning accuracy of the robot is estimated to be no better than 0.05- 0.1mm. Surface measurement is carried out using a Talysurf PGI surface profiler. Surface error correction is performed by removal of material from the surface through abrasion. The amount of material to be removed in each iteration is based on the measured surface error. Various material removal processes have been evaluated for surface correction. They involve using either bond abrasives or loose abrasives. A number of tool design options were investigated. Experiments were performed to evaluate the processes in terms of the material removal rate, process stability, etc. The current work on surface correction assumes a predetermined tool path. The feed rate along the tool path for surface correction is to be varied locally according to the required amount of material to be removed. Three algorithms were adopted / developed for the feed rate calculation. They are table look-up, discrete convolution and influence function offsetting method. Performance of these algorithms was evaluated through simulations. A free-form surface specimen of aluminium of about 25mm by 25mm across was used for the initial experiments. The surface was prepared by CNC surface machining before polishing. Through repeated measurement and surface correction, the peak to valley error of the surface was improved from 20 micron to 4.2 microns within a 20mm by 20mm surface region, and the surface roughness was improved from 496 nanometer to 13 nanometer.

    Research areas

  • Surfaces (Technology), Iterative methods (Mathematics)