January AMP_Digital

A D V A N C E D M A T E R I A L S & P R O C E S S E S | J A N U A R Y 2 0 1 9 1 8 Figure 2 and Table 1 show strong agreement between grain size distribu- tions and summary statistics. BSEmean grain size deviated from that of EBSD by only 2.7%. The standard deviations also show close agreement, but were not strongly considered in the comparison because they, together with maximum and minimum values, are more depen- dent on the particular fields imaged. A more thorough sampling is required to properly compare standard devia- tion as well as minimum and maximum statistics between the two. Most im- portantly, a 97.3% agreement between BSE and EBSD mean grain size delivers confidence in the ability to accurately and automatically perform grain size analysis from BSE images of the chal- lenging microstructure. IMPLICATIONS Accurate grain size measurement via BSE imaging rather than EBSD offers substantial cost savings and enhanced throughput. Table 2 presents an ap- proximate breakdown of time and ser- vice cost associated with each method. Mipar’s analysis time of five seconds per image was considered negligible, and is thus excluded from the cost breakdown. The true cost of grain size data ac- quisition is more complicated than indi- cated in Table 2 and likely varies among companies. However, it is not unreason- able to estimate that BSE imaging can offer savings in the thousands of dol- lars per sample compared with EBSD. Moreover, BSE imaging can conserva- tively offer 100 times the throughput of EBSD. Thus, while BSE imaging today could potentially process 240 samples in 24 hours with a direct cost of $4800, EBSD would require about four months to process the same volume, with a di- rect cost of $576,000. These benefits have been recog- nized for some time, but the challeng- es associated with automating grain detection from real-world SEM images forced engineers to resort to EBSD for grain sizing, despite the substantially higher cost. The ability to successful- ly overcome these challenges enabled Mipar software users to move to BSE grain sizing with significant cost savings and to process samples with greatly in- creased efficiency. OVERCOMING MEASUREMENT LIMITATIONS Powder and loose aggregate ma- terials are used in many engineering applications. Their physical properties determine powder flow characteristics, packing density, composite material properties, and the suitability of aggre- gates for various purposes. While this variety of properties determines aggre- gate behavior and suitability for use, many particle assays are restricted to particle size while other important in- formation is lost. Micrograph analysis retains more particle characteristics, enabling more accurate prediction of the aggregate’s behavior. TABLE 1 – GRAIN SIZE STATISTICS FROM SEM IMAGING METHODS [a] Backscattered electron (BSE) Electron backscatter diffraction (EBSD) Mean, µm 0.72 0.74 Standard deviation, µm 0.35 0.41 Minimum, µm 0.04 0.04 Maximum, µm 2.46 3.23 [a] Data frommeasurements from four random fields of view. TABLE 2 – TIME AND COSTS FOR BSE AND EBSD GRAIN SIZE ANALYSIS Backscattered electron (BSE) Electron backscatter diffraction (EBSD) Collection time [a] 50 s/sample 10 h/sample Samples per day [b] 240 2.4 Estimated SEM rate $200/h $200/h Collection cost per sample [a] $20/sample $2000/sample [a] Assuming five fields per sample. [b] Assuming 24 h SEM access and 5 min sample exchange time. Fig. 2 — (a) Outline of complete identified grains (red) fromBSE image; (b) outline of complete identified grains (white) from EBSD image; (c) grain size distribution from four BSE images; and (d) grain size distribution from four EBSD images (d) (c) (b) (a)