Production of a probe tip compensation method for reverse engineering free-form features

Reverse engineering is now a part of the modern engineering framework. It is used in the main to remanufacture products where documentation of the original design form is lost. Arm mounted Laser scanning instruments can collect large amounts of data rapidly at rates of 600K points per second, but at the expense of accuracy. Static coordinate measuring machines (CMM) can deliver more accurate measurements, but the process is slower with touch trigger probing data acquisition of the order of 25 points per minute. Prismatic objects that can be defined by equations (planar faces, cylinders etc.), are relatively easy to measure as their topology is well understood. Objects to be reverse engineered may be more difficult, for example castings formed from handmade patterns or forgings made from worn dies. An investigation has been carried out to determine the accuracy of an articulated arm mounted laser scanner measuring a diffuse and reflective reference planar target. Tactile probed data was gained from the targets to form a planar best fit reference. Laser scan data was taken of the targets at various angles and stand-off to their surface normals. Planar primitives were best fitted to this laser scan data. Scanner error was taken as the offset between the best fitted laser scanned plane and the tactile probed best fitted reference planes. The scanner systematic error was found to be most greatly influenced by scanner stand-off. Scanner systematic errors have been observed between ±0.070 mm. If surface microstructure produces speckle reflection, scan error becomes influenced not by this systematic error, but by the least squares residuals standard deviation of the planar feature best fit routine. Speckle reflection causes saturation of the CCD and laser scattering from the target microstructure causes high volumes of outliers. Two techniques have been developed to allow probe tip correction to be carried out on freeform features. Both techniques use laser scan data to represent the free form surface. This virtual representation allows the surface directly below a contact probe to be estimated and a compensation calculated in the probe approach direction. The second method uses the laser scan data representation of the target surface to determine the probe contact position with the target surface. Both methods were tested on virtual and actual scanned data for a 150.032 mm hemispherical form. The probe compensation in the probe approach direction gave adequate diameter errors of 0.018 mm for a 2 mm diameter probes, but 6 mm diameter probes gave diameter accuracies of 0.085 mm. Compensation in the surface contact normal direction produced results on a par with those expected from a static CMM probing prismatic parts for both the virtual and actual hemispherical form, producing diameter errors of 0.002 mm and 0.003 mm for 2 mm and 6 mm diameter probes respectively. Scan data density has proved a factor in the success of both compensation methods. It has been shown that sparse data reduces the accuracy of both methods. Prob tip correction establishing the probe contact point proved the most accurate and robust technique. This method was further validated with virtual data generated from a Bezier curve driven surface. Compensation errors from the compensated point to the virtual surface of 0.0015 mm were achieved. A value less than the CMM maximum permissible error (MPE) (0.0017 mm). Compensation accuracy was clearly seen to decrease with increasing probe diameter due to the basic trigonometry of the compensation method. Of the various data gathering techniques available, gaining laser scan data as a polygonal mesh in real time proved the most accurate and convenient as data cleaning of mesh data is a relatively straight forward process.