Statistical Engineering Division, ITLCarlos Beauchamp
Metallurgy Division, MSEL There are numerous industrial applications of measurements of the thickness of coatings: assessment of wear or build-up, compliance verification for anticorrosive coatings, precheck for successful assembly of coated parts, determination of optimal thicknesses for conductivity, insulation, or solderability. The earliest prototypes of the now most commonly used instruments for quick nondestructive evaluation of coating thicknesses for magnetic coatings on nonmagnetic substrates or nonmagnetic coatings on magnetic substrates were in fact developed by the Electrochemical Processing Group of the National Bureau of Standards in the 1940's. While implementations vary, all such instruments make use of the attenuation or distortion of a magnetic probe's field by the nonferric coating or substrate. And since that time, NIST has produced and maintained primary coating standards and disseminated secondary standards in the form of Standard Reference Materials (SRM's) for coating thickness assay.
Five years ago a major redesign and reissue effort for this family of SRM's was initiated. The new SRM's consist of sets of chromium-copper-plated steel coupons certified for coating thicknesses in the 2.5m-2000m range. Members of SED participated in all initial and ongoing design, analysis, and Quality Assurance aspects of the project. Work over the past year has concentrated mainly on the analysis of the calibration curve relating very homogeneous coupons chosen as primary standards, with thicknesses measured by atomic absorption, to subsets of the numerous secondary - SRM issue - standards measured by magnetic induction.
After clean-up of outliers and transcription
errors, and comparison of portions of the
global calibration curve to segments generated
using other methods (optical cross-section and
x-ray fluorescence), suitable descriptions - models -
of the calibration curve were sought. Global
parametric and nonparametric
versions of calibration models were
A global smoothing spline derivative
plotted on a logarithmic scale
gives strong indication of both how to segment
and how to parametrically model
With appropriate segment points chosen,
appropriate parametric models become
EXP[cubic], EXP[cubic], RECIP[quadratic].
For certification purposes,
uncertainties are quoted as confidence
bounds on the respective model-predicted
mean curves. For information purposes,
simple statistics on the actual fluctuation
of coating depth over the surface of the
coupon are also given.
Figure 27: Derivative of smoothing spline fit to global calibration curve plotted on a logarithmic scale.
Date created: 7/20/2001