Click for next page ( 8


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 7
between measured and target roundness for the #18 sured round was larger than the target round. For size class, which is the most prevalent size class. It Types 3 and 5 the average measured round was can also be observed from the table that the percent smaller than the target round value. The reason for roundness was underestimated for all size classes, this difference may be the different methods used for e.g., the measured percent round of 73.9% for the #20 separating the glass beads. The spiral separator and beads is smaller than the target roundness of 80%. roundometer were used in the preparation of the sam- This might be due to the difference between separa- ples but only the roundometer was used for testing of tion methods of the roundometer and the spiral sepa- the ILS samples. The measured values indicated that rator, which were both used in the preparation of the smallest repeatability and reproducibility coeffi- ILS samples. Similar to the previous observations, cients of variation correspond to the sieve with the the repeatability and reproducibility coefficients of largest number of beads. variation of roundness measurements correspond- ing to the size class with the largest amount of beads Analysis of Results from (#18 sieve) are smaller than those of other size classes. COM-A Measurements Type 5 Samples. Nine out of the 14 laboratories that A total of eight laboratories returned measure- conducted mechanical sieve analysis on the Type 5 ments by COM-A. The measurements included size samples also returned results on the roundness of the distribution and percent roundness. The roundness samples. The variability of percent round of the Type measurement was made using two parameters, b/l and 5 samples was determined after eliminating the out- SPHT, as explained in Table 3. The measurements lier data. All remaining data were re-analyzed accord- using COM-A were conducted following the instruc- ing to the E691 method to determine the repeatabil- tions provided by the COM-A manufacturer. The data ity and reproducibility statistics shown in Table 10. were analyzed to evaluate the precision and bias of A review of the statistics in the table indicates the each measured property and to compare these with the roundness of the Type 5 beads was underestimated precision and bias of the properties measured using for all size classes, e.g., the measured percent round the traditional methods and COM-B computerized of 83.4% for the #14 beads is smaller than the target optical method. roundness of 90%. This again could be due to the dif- ference between the methods of separating the beads Size Measurements using the roundometer and the spiral separator. The The size distributions of the three sample types repeatability and reproducibility coefficients of vari- were measured using two-dimensional (2-D) analy- ation of the Type 5 beads are relatively low for all sis of the images of the glass beads passing through size classes. This might indicate better control over the COM-A measurement unit. the roundness determination of larger beads than of smaller beads. Type 1 Samples. The precision estimates for size dis- tribution of Type 1 specimens were determined after Summary of Percent Round by the Roundometer. eliminating the outlier data. All remaining data were The analysis of the roundometer percent round data re-analyzed according to the E691 method to deter- suggests that for Type 1 samples, the average mea- mine the repeatability and reproducibility statistics Table 10 Roundness statistics of Type 5 samples using roundometer, target roundness of 90% No. Measured % Repeatability Reproducibility Sieve of Round, Sizes Labs Average STD, % CV, % STD, % CV, % #12 7 86.1 3.1 3.6 5.0 5.8 #14 8 83.4 1.8 2.1 4.5 5.3 #16 7 87.0 2.1 2.4 4.1 4.7 #18 7 87.3 2.9 3.4 4.8 5.4 7

OCR for page 7
Table 11 Percent retained statistics of Type 1 samples using COM-A No. Target Measured Repeatability Reproducibility Sieve of % % Retained, Sizes Labs Retained Average STD, % CV, % STD, % CV, % #30 6 5 4.5 0.3 5.6 0.3 7.1 #50 8 50 46.4 1.4 3.0 3.0 6.5 #100 8 45 48.6 1.6 3.3 3.3 6.7 shown in Table 11. As indicated in Table 11, the mea- retained value of 57.2% compares relatively well with sured mass percents retained are in relatively good the target retained value of 55%. The repeatability and agreement with the target values. For example, the reproducibility coefficients of variation for the #18 average retained of 46.4% for the #50 sieve size is sieve are smaller than those for the other sieves. compared with the target retained of 50%. Similar to the mechanical sieve analysis of Type 1 samples, the Type 5 Samples. The precision estimates for size dis- repeatability and reproducibility coefficients of varia- tribution of the Type 5 specimens were determined tion of all class sizes are relatively small. after eliminating the outlier data. All remaining data were re-analyzed according to the E691 method to Type 3 Samples. The precision estimates for size dis- determine the repeatability and reproducibility sta- tribution of Type 3 specimens were determined after tistics shown in Table 13. A very good agreement eliminating the outlier data. All remaining data were between the measured and target percent retained is re-analyzed according to the E691 method to deter- observed for the #14 size class, which has the largest mine the repeatability and reproducibility statistics number of beads. The measured retained value of shown in Table 12. A comparison of the measured and 55.5% compares very well with the target retained target percent retained in Table 12 indicates a good value of 55%. It is also indicated from the table that agreement between the measured percent retained the smallest repeatability and reproducibility coeffi- and the target percent retained in the #18 size class, cients of variation for the percent retained correspond which has the largest number of beads. The measured to #14 sieve. Table 12 Percent retained statistics of Type 3 samples using COM-A No. Target Measured Repeatability Reproducibility Sieve of % % Retained, Sizes Labs Retained Average STD, % CV, % STD, % CV, % #16 7 5 10.3 0.4 4.0 3.4 32.6 #18 8 55 57.2 0.7 1.2 2.1 3.7 #20 8 35 26.8 0.8 3.0 1.8 6.7 #25 8 5 4.6 0.7 15.4 0.9 19.6 Table 13 Percent retained statistics of Type 5 samples using COM-A No. Target Measured Repeatability Reproducibility Sieve of % % Retained, Sizes Labs Retained Average STD, % CV, % STD, % CV, % #12 8 5 6.9 0.7 10.2 1.4 19.6 #14 7 55 55.5 0.9 1.5 1.2 2.1 #16 7 35 30.1 1.7 5.7 2.1 6.9 #18 7 5 6.5 0.3 4.2 0.7 10.3 8

OCR for page 7
Table 14 Statistics of percent round of Type 1 samples using SPHT parameter of COM-A, target 70% No. Measured % Repeatability Reproducibility Sieve of Round, Sizes Labs Average STD, % CV, % STD, % CV, % #30 4 80.8 1.9 2.3 5.8 7.2 #50 3 84.5 0.7 0.8 1.5 1.8 #100 4 84.8 2.5 2.9 2.2 2.6 Summary of Percent Retained by COM-A. The analy- cantly larger than the target roundness of 70%. This sis of COM-A mass percent retained data suggests that suggests that the SPHT threshold value may be unsuit- the measured percent retained in the size class with the able because it allows non-round beads to be classified largest number of beads provides closest agreement as round. Despite the inaccuracy of the SPHT param- with the target value. Looking at the percent retained eter, both the repeatability and reproducibility coef- in the size classes with the largest number of beads ficients of variation of the SPHT parameter are rela- in the Type 1, Type 3, and Type 5 samples indicates tively small for all size classes. that both accuracy and precision of measurements improved with the coarseness of the glass bead types. Type 3 Samples. The precision estimates for the mass percent roundness of the Type 3 specimens were Roundness Measurements determined after eliminating the outlier data. All Using the SPHT Parameter remaining data were re-analyzed according to the A total of four laboratories reported the percent E691 method to determine the repeatability and roundness of the three sample types measured by the reproducibility statistics shown in Table 15. A review SPHT parameter (see Table 3) using the COM-A of the statistics in Table 15 indicates that the percent device. A threshold value of 0.9 was used for round- round of the size classes with the most number of ness determination using the SPHT parameter, mean- beads (#18 and #20) was overestimated by the SPHT ing that any particle with an SPHT value of 0.9 and parameter. The measured percent round of 86.3% for above was considered to be round. the #18 beads is compared with the target roundness of 80%. This might indicate that the SPHT threshold Type 1 Samples. The precision estimates for the value is too low, which allows non-round beads to mass percent roundness of the Type 1 specimens be classified as round. The repeatability and repro- were determined after eliminating the outlier data. ducibility coefficients of variation of the roundness All remaining data were re-analyzed according to the data obtained based on the SPHT parameter are rela- E691 method to determine the repeatability and repro- tively small for all size classes. ducibility statistics shown in Table 14. Table 14 indi- cates that the roundness of all size classes are over- Type 5 Samples. Based on outlier analysis, no data estimated by the SPHT parameter, e.g., the measured were eliminated from the determination of the preci- percent round of 84.5% for the #50 beads is signifi- sion estimates for the mass percent roundness of the Table 15 Statistics of percent round of Type 3 samples using SPHT parameter of COM-A, target 80% No. Measured % Repeatability Reproducibility Sieve of Round, Sizes Labs Average STD, % CV, % STD, % CV, % #16 4 77.7 2.6 3.4 2.9 3.7 #18 4 86.3 0.9 1.0 1.0 1.2 #20 3 92.1 0.4 0.4 0.8 0.8 #25 4 85.1 3.4 4.0 4.1 4.9 9

OCR for page 7
Table 16 Statistics of percent round of Type 5 samples using SPHT parameter of COM-A, target 90% No. Measured % Repeatability Reproducibility Sieve of Round, Sizes Labs Average STD, % CV, % STD, % CV, % #12 4 88.5 1.0 1.1 2.6 2.9 #14 4 92.1 0.6 0.7 1.2 1.4 #16 4 91.2 1.7 1.8 5.7 6.2 #18 4 92.1 0.7 0.7 0.7 0.7 Type 5 specimens. The computed repeatability and cally, a threshold value of 0.9 appears acceptable for reproducibility statistics of the Type 5 sample round- Type 5 beads, but it did not measure the intended ness are shown in Table 16. A review of the statistics roundness of Type 1 and Type 3 beads accurately. in Table 16 indicates that the roundness of all size classes was measured reasonably correctly by the Roundness Measurements Using b/l Parameter SPHT parameter. The measured percent round of Eight laboratories reported the percent roundness 92.1% for the #14 beads agrees well with the target of the three sample types measured by the COM-A b/l roundness of 90%. In addition, a review of the vari- parameter. The threshold value for measuring per- ability values in Table 16 indicates that both the cent round by the b/l parameter was 0.83, meaning repeatability and reproducibility coefficients of vari- that any particle with b/l > 0.83 was considered to be ation of the SPHT parameter are very small for all size round. classes. This suggests that (1) the threshold value of 0.9 is more appropriate for larger beads than for Type 1 Samples. The variability of percent round smaller beads and (2) a higher threshold value is of Type 1 specimens was determined after elimi- needed for more accurate measurements of the round- nating the outlier data. All remaining data were re- ness of smaller glass beads. analyzed according to the E691 method to determine the repeatability and reproducibility statistics shown Summary of Percent Round by COM-A SPHT. The in Table 17. A review of the statistics in Table 17 analysis of mass percent round using the COM-A indicates that the roundness of all size classes of the SPHT parameter indicates that the measured percent Type 1 samples is overestimated by the b/l param- round in the most prevalent size classes provided eter. The measured percent round of 78.6% for the closest agreement with the target value. However, the #50 beads is larger than the target roundness of 70%. level of agreement between measured and target per- This difference may result from an artifact of the cent round differs for the Type 1, Type 3, and Type 5 2-D image analyses. A single or even several 2-D samples. This suggests that the threshold value for projections of a non-spherical object cannot fully all glass bead types is not the same and would need capture its 3-D shape, which would tend to bias the adjustment according to the glass bead type. Specifi- percent round results. This is also an indication that Table 17 Statistics of percent round of Type 1 samples using b/l parameter of COM-A, target 70% No. Measured % Repeatability Reproducibility Sieve of Round, Sizes Labs Average STD, % CV, % STD, % CV, % #30 8 76.8 2.5 3.2 6.1 8.0 #50 6 78.6 1.2 1.6 1.7 2.1 #100 7 80.9 3.1 3.8 3.5 4.3 10

OCR for page 7
Table 18 Statistics of percent round of Type 3 samples using b/l parameter of COM-A, target 80% No. Measured % Repeatability Reproducibility Sieve of Round, Sizes Labs Average STD, % CV, % STD, % CV, % #16 8 66.5 2.2 3.3 3.4 5.0 #18 8 80.1 0.9 1.1 1.1 1.4 #20 7 84.3 0.9 1.0 1.1 1.3 #25 8 76.6 4.7 6.1 9.5 12.4 the threshold value for the b/l parameter is not large of the most prevalent size classes of Type 3 glass enough to eliminate all particles considered to be beads. non-round by the roundometer. Similar to the obser- vations made on the roundometer data, both the Type 5 Samples. The variability of percent round of repeatability and reproducibility coefficients of vari- Type 5 specimens was determined after eliminating ation of the b/l parameter are significantly smaller for the outlier data. All remaining data were re-analyzed the #50 beads, which are the most prevalent size in the according to the E691 method to determine the Type 1 samples. repeatability and reproducibility statistics shown in Table 19. A review of the statistics in Table 19 indi- Type 3 Samples. The variability of percent round of cates that the measured roundness of the #14 and Type 3 specimens was determined after eliminating #16 glass beads, which are the most prevalent sizes, the outlier data. All remaining data were re-analyzed agrees very well with the target roundness of 90%. according to the E691 method to determine the The repeatability and reproducibility coefficients repeatability and reproducibility statistics shown in of variation corresponding to the #14 and #16 size Table 18. A review of the statistics in Table 18 indi- classes are also very small. This indicates that b/l is a cates that the measured percent round of the #18 reliable parameter for measuring the roundness of the glass beads (80.1%), which is the most prevalent most prevalent size classes of Type 5 glass beads. size, agrees very well with the target roundness of 80%. The percent round of the #20 sieve is slightly Summary of Percent Round by COM-A b/l. The analy- overestimated, which may also be due to 2-D image sis of mass percent round of COM-A b/l data indicates analysis artifacts as noted in the previous section. The that the b/l parameter captured the roundness of results in Table 18 further indicate that the repeata- Type 3 and Type 5 glass beads very well but over- bility and reproducibility coefficients of variation estimated the roundness of Type 1 beads. This indi- of percent round according to the b/l parameter for cates that the threshold value for b/l should not be the the #18 and #20 beads, which are the most preva- same for all glass bead types. While the threshold value lent sizes, are very small. This indicates that b/l is of 0.83 seems adequate for Type 3 and Type 5 glass a reliable parameter for measuring the roundness beads, it may need to be increased for Type 1 beads. Table 19 Statistics of percent round of Type 5 samples using b/l parameter of COM-A, target 90% No. Measured % Repeatability Reproducibility Sieve of Round, Sizes Labs Average STD, % CV, % STD, % CV, % #12 8 80.8 1.2 1.5 3.0 3.7 #14 8 89.2 1.1 1.2 1.5 1.7 #16 7 91.4 1.9 2.1 2.3 2.6 #18 8 89.8 1.1 1.2 1.4 1.5 11

OCR for page 7
D10, D50, and D90 Measurements Table 21 Comparison of measured and target particle sizes of Type 3 beads for 10%, 50%, and 90% passing The accuracy of COM-A in measuring the size distribution of glass beads was evaluated. For this Sieve Target Particle purpose, the three diameters where 10%, 50%, and No. of Size Percent DiameterXcmin Percent 90% of the particles, by mass, are smaller than these Labs (mm) Passing (mm) Smaller diameters (D10, D50, D90) were compared with 7 0.85 5 0.89 10 the sieve sizes and the mass percent passing that 7 1.00 40 1.07 50 were used to prepare the bead samples. Tables 20, 21, 7 1.18 95 1.18 90 and 22 provide the measured diameters correspond- ing to 10%, 50%, and 90% of particles having a diam- eter less than the given diameter. These tables also instrument has measured the size distribution of the provide the sieve sizes and the percent passing of Type 3 samples reasonably well. The percent smaller Types 1, 3, and 5 glass beads prepared in this study. for D90 has an expected value of 95% compared to a measured value of 90%, but the percent smaller Type 1 Samples. Column 4 of Table 20 shows the than D10 and D50 are reasonable. For example, 40% values of D10, D50, and D90 of Type 1 samples of the beads were prepared to pass through a 1.0 mm according to COM-A data. The results in Column 4 opening (#18 sieve) and the COM-A predicted that are averages of COM-A measurements received from 50% of the beads have diameter smaller than 1.07 mm, seven laboratories. The comparison of the measured which is reasonable agreement. and target values of particle size with respect to per- cent smaller and percent passing values indicates that Type 5 Samples. Column 4 of Table 22 provides the the COM-A instrument has measured the size distri- sizes in millimeters for D10, D50, and D90, as deter- bution of the Type 1 samples reasonably well. For mined from the COM-A results. The second and third example, bead samples were prepared to have 95% columns provide the sieve openings and the cor- passing a 0.6 mm sieve opening (#30 sieve) and the responding percent passing used in preparing the COM-A data measured that 90% of the beads are Type 5 bead samples. The comparison of the mea- smaller than 0.53 mm. The same logic is applied to sured and target values of particle size with respect to other sieve sizes of Type 1 samples, which indicates the percent smaller and percent passing indicates that that the COM-A device, using the width (b or Xcmin) the COM-A device has measured the size distribution parameter, has measured the size distribution of of the Type 5 samples reasonably well. For exam- Type 1 beads reasonably well. ple, 40% of the beads were prepared to pass through a 1.40 mm opening (#14 sieve); the COM-A result is Type 3 Samples. Column 4 of Table 21 provides the that 50% of the beads by mass have diameters smaller values of D10, D50, and D90, in terms of the param- than 1.49 mm. eter Xcmin, for the Type 3 samples; Columns 2 and 3 provide the sieve openings and the corresponding Summary of D10, D50, and D90 Measurements by percent passing used in preparing the Type 3 sam- COM-A. The analysis of D10, D50, and D90 data ples. The comparison of the measured and target measured by COM-A indicated that COM-A mea- values of particle size with respect to percent smaller sured the size distribution of the three types of glass and percent passing values indicates that the COM-A bead samples reasonably well. Only one out of nine Table 20 Comparison of measured and target particle Table 22 Comparison of measured and target particle sizes of Type 1 beads for 10%, 50%, and 90% passing sizes of Type 5 beads for 10%, 50%, and 90% passing Sieve Target Particle Sieve Target Particle No. of Size Percent DiameterXcmin Percent No. of Size Percent DiameterXcmin Percent Labs (mm) Passing (mm) Smaller Labs (mm) Passing (mm) Smaller 7 0.15 0 0.20 10 7 1.18 5 1.21 10 7 0.30 45 0.31 50 7 1.40 40 1.49 50 7 0.60 95 0.53 90 7 1.70 95 1.66 90 12