Certification of New Standard Reference Material 2806b Medium Test Dust in Hydraulic Fluid

Abstract

A new material has been certified to become Standard Reference Material (SRM) 2806b - Medium Test Dust in Hydraulic Fluid. SRM 2806b consists of trace polydisperse, irregularly shaped mineral dust particles suspended in hydraulic fluid. The certified values of SRM 2806b are the projected area circular-equivalent diameters of the collected dust particles from the hydraulic fluid. The dimensional measurements were determined from the area of the collected dust particles using images obtained from automated scanning electron microscopy (SEM) followed by image analysis. An automated SEM and an automated image analysis software allowed the processing of over 29 million particles. The dimensional calibration of the SEM images (actual length per pixel and thus the actual projected diameters) are traceable to the NIST Line Scale Interferometer (LSI) through a NIST calibrated Geller MRS-4XY pitch standard. The certified diameters are correlated with the numeric concentration of particles greater than each diameter, referred to as the cumulative number size distribution. SRM 2806b is intended to be used to calibrate liquid-borne optical particle counters in conjunction with the reference method ISO 11171:2010.

Keywords: SRM; calibration material; diameter; image analysis; liquid optical particle counter; particle; reference; scanning electron microscopy (SEM); standard.

Fig. 1.

Backscatter SEM image of medium test dust. Field-of-view is 250 μm.

Fig. 2.

High-resolution image of the Geller standard – 8192 × 8192 pixels, 66 MB. The region calibrated was over the 50 μm spacing both horizontally and vertically.

Fig. 3.

Set of 517 images collected on a filter at the lowest (2.0 mm) magnification. Double edges indicate field overlap.

Fig. 4.

Filter diagram showing 2.0 mm fields (large squares) and within them, the randomly selected (with blocked coverage constraints) 0.25 mm fields (small squares) over the entire filter.

Fig. 5.

Tiled SEM image of the entire filter. Filtration area is inside the manually drawn red circle.

Fig. 6.

Diagram showing image overlap.

Fig. 7.

Diagram that illustrates assignment of particles in the overlap region to image A or image B. Particles are shown as circles with centroids represented as black dots.

Fig. 8a.

Threshold for point ‘a’ in Fig. 9, selected pixels are shown in red. Threshold is too high: because some low intensity (dim) particles are not selected (red) and most particles are not completely selected (white holes or edges).

Fig. 8b.

Threshold for point ‘b’ in Fig. 9. Most particles completely selected.

Fig. 8c.

Threshold for point ‘c’ in Fig. 9. Optimal point for objects and background. Even dimmest particles selected, but background features are not.

Fig. 8d.

Threshold for point ‘d’ in Fig. 9. Threshold too low: too many background pixels selected, which are starting to form clusters that mimic particles.

Fig. 8e.

Threshold for point ‘e’ in Fig. 9. Threshold far too low: particles and background all fused together.

Fig. 9.

Plot of particle area fraction P / N vs. pixels intensity threshold. Corresponding image samples above in Figs. 8a–e. Appropriate threshold is at point c.

Fig. 10.

Horizontal NIST calibrated area of standard: left half of the image (between horizontal dashed lines—see Fig. 2). Measurements on the left extended to the right half of image. Green—nominal 100 μm distances. Red letters: NIST-traceable LSI calibrated features, distances measured—from feature a to features b through i. Blue letters: secondary features.

Fig. 11.

Disc (outlined in blue) represented by pixels. The thermal color scheme provided shows what degree the pixel is covered by the disc.

Fig. 12.

Percent difference in area in the low magnification image for single particles vs. particle area in pixels for low magnification images.

Fig. 13.

Plot of cumulative number of particles per milliliter of hydraulic fluid greater than the specified projected area diameter. The expanded uncertainties (U) are shown in the plot. The power law distribution shown in the plot is frequently characteristic of natural fracturing of materials.

Fig. 14.

Monte Carlo simulation used to calculate the combined standard and expanded uncertainties for the main two components, digitization and random sampling. A fiducial distribution is one which is constructed from various considerations so as to reflect the “likely” distribution of errors from a component in a measurement process. For component 1 (digitization error) geometric arguments suggest a U-shaped error distribution as appropriate—the well-known statistical beta distribution was utilized to formally model such U-shapedness. For component 2 (random measurment error), the error distribution was likely symmetric and unimodal, and so the usual Gaussian distribution was assumed. Monte Carlo simulation was then used to form the propagated particle size distribution from which the combined standard and expanded uncertainties were calculated.