Characteristics, Accuracy and Reverification of Robotised Articulated Arm CMMs

VDI article 2617 1 specifies characteristics to describe the accuracy of articulated arm coordinate measuring machines (AACMMs) and outlines procedures for checking them. However the VDI prescription was written with a former generation of machines in mind: manual arms exploiting traditional touch probe technologies. Recent advances in metrology have given rise to noncontact laser scanning tools and robotic automation of articulated arms – technologies which are not adequately characterised using the VDI specification. In this paper we examine the “guidelines” presented in VDI 2617, finding many of them to be ambiguous and open to interpretation, with some tests appearing even to be optional. The engineer is left significant flexibility in the execution of the test procedures and the manufacturer is free to specify many of the test parameters. Such flexibility renders the VDI tests of limited value and the results can be misleading. We illustrate, with examples using the Nikon RCA, how a liberal interpretation of the VDI guidelines can significantly improve accuracy characterisation and suggest ways in which to mitigate this problem. We propose a series of stringent tests and revised definitions, in the same vein as VDI 2617 and similar US standards, to clarify the accuracy characterisation process. The revised methodology includes modified acceptance and reverification tests which aim to accommodate emerging technologies, laser scanning devices in particular, while maintaining the spirit of the existing and established standards. We seek to supply robust re-definitions for the accepted terms “zero point” and “useful arm length”, presupposing nothing about the geometry of the measuring device. We also identify a source of error unique to robotised AACMMs employing laser scanners – the forwardreverse pass error. We show how eliminating this error significantly improves the repeatability of a device and propose a novel approach to the testing of probing error based on statistical uncertainty.