Statistical Engineering Division, CAML
Precision Engineering Division, MEL
Factory Automation Systems Division, MEL
Applied and Computational Mathematics Division, CAML
Department of Mechanical Engineering, George Washington University
U.S. industry uses more than 20,000 coordinate measuring machines (CMMs). A CMM helps determine if a part conforms to design specifications by measuring the coordinates of a sample of locations on the part surface. Currently, there is no rigorous methodology to determine the accuracy and the precision of the measurements from a CMM. Consequently, CMMs are considered uncalibrated and not traceable to the SI according to ISO definitions. Developing calibration methodology for CMMs is necessary for U.S. companies to trade internationally. Additionally, the methodology would promote improvement in quality and efficiency through better determination of part dimensions. As part of a NIST competency project, SED scientists play an active role in a cross-disciplinary group that is making significant progress towards solving this problem. So far, the group has concentrated on understanding and modeling the CMM probe, which is the largest source of error in the measuring process. The probe is the component of the CMM that senses the contact with the part being measured. The most common class of probes has a construction that leads to pronounced systematic effect in the CMM measurements. Through a large modeling effort, the group has produced a reliable model that allows for real-time correction of measurements. The result is an improved measurement system without significant added costs.
SED scientists have developed an uncertainty procedure for use with the model. The procedure addresses the overall project goal of traceability. Three independent sources make up the overall uncertainty. The largest source derives from the estimation of the probe model. The upper left plot in the accompanying figure shows a dense sample of points collected on a circle. The systematic effect of the probe appears on this exaggerated scale. The model estimated from the dense data, shown in the upper right plot of the figure, is considered to be near-ideal. However, in practice it will not be feasible to collect so many points in a manufacturing environment, and the model must be estimated based on a small sample of points. The remaining figures show how the model estimates vary because of this sampling variation.
Figure 7: Probe Model Estimates
Date created: 7/20/2001