An intercomparison measurement has been carried out by 12 manufacturers of form measurement equipment, industrial users and national metrology institutes in order to document the present day state of technical form calibrations at an accuracy level of approximately 0.1 um and to point out ways towards further harmonization.
2 cylindrical transfer standards have been measured by all participants of the comparison to determine the deviations from roundness, straightness, parallelism and cylindricity. The standards were chosen to have some roundness and straightness traces with form deviations comparable to the guide errors of the used instruments (up to 0.5 um) as well as other traces with characteristic form deviation features (up to 15 um), allowing investigation of the amplification and transmission properties of the instruments.
Satisfactory accordance of the participants' results has been observed with respect to single measuring lines (roundness and straightness), whereas large variations have been found concerning the results on parallelism and cylindricity calibrations, where one has to combine the position and form deviations of several measuring lines to calculate the investigated quantities. It is recommended that future efforts towards a higher degree of harmonization of form measurement technique should concentrate on the different calibration techniques of the transducer systems and the different software filter characteristics.
An intercomparison of high accuracy roundness measurements has been carried out. The national standards laboratories of 10 countries participated in this exercise. The circulated artefacts were a hemispherical glass standard, with nominal roundness deviation of 0.04 um, and a steel disc with nominal roundness deviations of 0.1 and 0.4 um.
This exercise involved the measurements of 8 roundness profiles, 2 on the hemisphere and 6 on the disc. The eccentricity between external and internal surfaces of the disc was determined, and the effect of various measuring conditions was evaluated. The overall results show a satisfying agreement for almost all laboratories, but room for further improvement still exists.
The uncertainty limit was mainly due to the spindle errors of the roundness instruments used in each laboratory. These instruments were proven to be of a good quality at 0.1 um level. Most significant results were given by those laboratories where error separation techniques were available.
The measurements of roundnesses are amongst the most important quality control tasks undertaken by manufacturing industry. However, it is recognised that significant discrepancies in the assessment of the roundness of work pieces exist due mainly to insufficient standardisations of measurement procedures and conditions. It is the aim of the project to understand the metrological problems which lie behind the lack of standards in order to achieve a harmonisation of roundness measurements throughout the Community.
The project consisted of two intercomparisons:
a) The roundness of two highly accurate transfer standards, a hemispherical glass roundness standard and a circular master was measured by national metrology institutes (NMI's). Four laboratories applying error separation tecniques agreed within +/- 25 nm. In some instances their 95% confidence intervals (CI) (up to 25 nm) did not overlap indicating that there are systematic errors. Laboratories without such techniques agreed within +/- 50 nm the 95% CI being up to 125 nm. For these laboratories, noise caused problems when higher frequencies were not filtered. Using a hemisphere, systematic differences (10 nm) depending on the stylus direction (normal to the axis or to the surface) are apparent.
b) Roundness, straightness, parallelity and cyclindricity on an industrial level were measured by NMI's and industry. A cyclinder with form deviations and two flats for the calibration of the transducers was circulated as well as a waviness standard. Excluding outliers, the roundness measurements were in better agreement than expectd. The maximum difference between the dynamic calibrations of the transducers when using the flats was 1.3 um corresponding to 10%. The waviness standards are a promising alternative which should be further investigated. Filter conditions have a significant influence on the results and need to be harmonised. The measurement and calculation of profiles which are necessary to determine composed forms (parallelity, cylindricity were not statisfactory and need to be improved.
Funding SchemeCSC - Cost-sharing contracts