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 59
59
4.00 Binder
C
D
3.75 E
F
Tensile Strength (MPa)
H
3.50 I
J
K
M
3.25
N
O
3.00
2.75
2.50 Modified binders are shown with dashed connect lines
110 130 150 170
Compaction Temperature (C)
Figure 41. Interaction plot for compaction temperature and binder
on tensile strengths at 10°C.
it was not possible to make general conclusions or observations observed in temperatures needed to achieve the baseline coat-
about trends regarding the effect of compaction temperature. ing percentage with the bucket mixer. Binders that had the
largest residuals were O, C, B, and I. The temperatures to
achieve good coating for Binders O and C were much higher
Correlation of Mixing and
than predicted by the SSF method and for B and I coating was
Compaction Temperatures
achieved at temperatures much lower than predicted by the
A key part of this research was to compare the predicted SSF method.
mixing and compaction temperatures from the candidate Figure 43 shows a similar set of plots of SSF mixing temper-
binder tests with the results of the mixture experiments. This atures versus the mixing temperatures to achieve 89% coating
section presents that analysis using correlations performed with the pugmill mixer. The regression statistics are somewhat
with MINITAB release 15 statistical software. The regressions poorer compared with those with the bucket mixer. For this
are based on average values from replicate mix tests and binder data set, Binder E was considered as a possible outlier since the
tests. Outputs included graphical plots of the data, least- predicted temperature to achieve the baseline percentage was
squares linear regression equations, 95% confidence intervals 222°F (106°C), which seems quite low and is outside of the
for the regressions, the residual error regression S, the coeffi- experimental range.
cient of determination R2, and the adjusted R2. An ANOVA Figure 44 shows the correlations of mixing temperatures
table was also generated for each regression, which includes the from the Phase Angle method with the coating test results
observed significance level (P-value) for the regression equa- using the bucket mixer. As noted above, the bucket mixer coat-
tions. For several correlations, one or two data points were ing test results for Binder H were not considered valid, so these
identified as suspected outliers. Discussion of the outliers is data were removed and the correlations were performed again.
presented as part of the analysis associated with the particular The statistics for the regressions in these two plots are poorer
correlation. In each case where outlier data is suspected, the than for the SSF method. As with the SSF correlations, the
correlations are provided both with and without the question- binders that have large residuals include Binders O, C, I, and
able data. B. Binder M also has a large residual in the Phase Anglebucket
Figure 42 shows the correlations of the mixing tempera- mixer coating test correlations. This is the binder that includes
tures from the Steady Shear Flow (SSF) method with the mix- the SasobitŪ wax. The large residual with this binder could
ing temperatures to achieve 98% coating using the bucket indicate that the Phase Angle method is not a good predictor
mixer. The bucket mixer coating test results for Binder H was of mixing temperatures for binders with SasobitŪ, perhaps
not considered reasonable as the predicted temperature to because the Phase Angle measurements are made at below
achieve the baseline coating percentage was outside of the temperatures where the SasobitŪ wax melts. The poor correla-
experimental range. The regression statistics indicate the SSF tions with O, C, I, and B for both the SSF and the Phase Angle
mixing temperature explains only about 40% of the variation methods is evidence that the problematic data likely arise
OCR for page 60
60
Bucket Mix T = - 70.9 + 1.259 SSF Mix T
425 Regression
H
95% CI
400 S 29.9023
R-Sq 43.5%
R-Sq(adj) 38.4%
375
C
Bucket Mix T
O G
350
N
325 F
B
300 K J
D
E M
I
275
250
280 290 300 310 320 330 340
SSF Mix T
(a)
Bucket Mix T = 36.9 + 0.8928 SSF Mix T
375 Regression
95% CI
C
S 22.1202
350 O G R-Sq 40.8%
R-Sq(adj) 34.9%
N
Bucket Mix T
325 F
B
300 K J
D
E M
I
275
250
280 290 300 310 320 330 340
SSF Mix T
(b)
Figure 42. Correlations of the SSF mixing temperature with
the bucket mixer temperature for 98% coating: (a) all data;
(b) excludes Binder H.
from the bucket mixer coating results rather than the can- Overall, the correlations between the mixing temperatures
didate methods for determining mixing and compaction from the candidate methods and the coating test results are
temperatures. fairly weak, generally with R2 values in the range of 30 to 40%.
Figure 45 shows the correlations of the Phase Angle method However, the lack of strength of the correlations is likely due
predicted mixing temperatures with the temperatures to more to the coating test results that are based on curve-fitting
achieve the baseline coating percentage in the pugmill mixer. through data from subjective measurements that lack good
As discussed with the SSF methodpugmill mixer correlation, repeatability.
the coating test results for Binder E in the pugmill were not Figure 46 and Figure 47 show correlations between the can-
considered valid. The regression statistics between the Phase didate methods mixing temperatures and the midpoints of the
Angle method and the pugmill mixer coating results are slightly mixing temperature range recommended by the binder suppli-
better than for the SSF method. A summary of correlation ers. The correlation between the SSF method mixing tempera-
statistics is shown in Table 36. ture and the suppliers' recommended mixing temperatures is
OCR for page 61
61
Pugmill Mix T = - 71.5 + 1.244 SSF Mix T
400 Regression
95% CI
B
N S 34.8156
R-Sq 35.7%
350 M R-Sq(adj) 29.9%
I G
C
Pugmill Mix T
J O H
K F
300
D
250
E
200
280 290 300 310 320 330 340
SSF Mix T
(a)
Pugmill Mix T_1 = 28.4 + 0.9426 SSF Mix T
Regression
B
375 95% CI
N S 27.3356
R-Sq 34.6%
350 M R-Sq(adj) 28.1%
Pugmill Mix T_1
I G
C
325 J O H
K F
300
275
D
250
280 290 300 310 320 330 340
SSF Mix T
(b)
Figure 43. Correlations of the SSF mixing temperature with
the pugmill mixer temperature for 89% coating: (a) all data;
(b) excludes Binder E.
quite reasonable, with an R2 of 70%. The correlation of the Figure 48 and Figure 49 show the correlations between the
Phase Angle method mixing temperatures with the suppliers' workability tests and the results of the candidate methods for
recommended mixing temperatures is very weak unless the determining mixing and compaction temperatures. For these
results from Binder M are removed. The issue with the Saso- correlations, the temperatures midway between the mixing
bitŪ wax in Binder M was noted previously. With this data and compaction temperatures from the candidate methods
point removed, the correlation statistics improve consider- were used as the independent variable. Although the correla-
ably, although not as strong as with SSF method. However, it tions are weak as indicated by the low correlation coefficients,
is also important to note that the regression equation between the regressions were statistically significant ( =0.05). No data
the phase angle method and producers' recommendations were excluded from these correlations. However, most of the
was slightly closer to the line of equality than for the SSF scatter is likely due to poor precision of the workability tests.
method. Overall, these correlations show that both methods Despite numerous attempts to improve the workability equip-
provide results generally consistent with field experience and ment and test method during this study, there is considerable
therefore pass a test of reasonableness. doubt about the validity of the test as an indicator of binder
OCR for page 62
62
Bucket Mix T = - 48.5 + 1.153 Phase Angle Mix T
425 Regression
H
95% CI
400 S 34.8178
R-Sq 23.5%
R-Sq(adj) 16.5%
375
C
Bucket Mix T
O G
350
N
325 F
B
300 KJ D
E M
I
275
250
300 310 320 330 340
Phase Angle Mix T
(a)
Bucket Mix T_1 = 58.5 + 0.7970 Phase Angle Mix T
C
Regression
360
95% CI
O G S 25.3382
R-Sq 22.4%
340 N R-Sq(adj) 14.6%
Bucket Mix T_1
F
320
B
300 J
K D
E M
I
280
260
300 310 320 330 340
Phase Angle Mix T
(b)
Figure 44. Correlation of Phase Angle Method mixing temperature
with the bucket mixer temperature for 98% coating: (a) all data;
(b) excludes Binder H.
stiffness on mix workability. The average temperature differ- tests: Binders I, J, and N. It is clear from the correlation graphs
ence for the duplicate runs on all of the workability tests was that the SGC compaction test temperatures for I and J are well
26°F, which is a similar magnitude as the residuals for the cor- above the reasonable range for these binders. Binder N is a
relations. heavily modified binder that was expected to require a rela-
Correlations between the compaction temperatures pre- tively high temperature to achieve the baseline density level.
dicted by the candidate methods with the results of the com- Its compaction experiment results were just above the highest
paction tests are shown in Figure 50 and Figure 51. Excluding test temperature used in the compaction experiment.
the data for Binders I and J, both of the candidate methods' A summary of the regression statistics for the correlations
compaction temperatures correlate well with the results from between the candidate methods and mix test results is shown
the mix compaction tests. Three binders had compaction test in Table 36. Also shown are the key statistics from the corre-
results that were outside of the experimental range for the lations with the respective midpoints of the binder producers'
OCR for page 63
63
Pugmill Mix T = - 188.0 + 1.572 Phase Angle Mix T
400 Regression
95% CI
B
N S 34.5577
R-Sq 36.7%
350 M R-Sq(adj) 30.9%
I G
C
Pugmill Mix T
J O H
K F
300
D
250
E
200
300 310 320 330 340
Phase Angle Mix T
(a)
Pugmill Mix T_1 = - 86.2 + 1.275 Phase Angle Mix T
B Regression
375 95% CI
N
S 25.7628
R-Sq 41.9%
350 M R-Sq(adj) 36.1%
I G
Pugmill Mix T_1
C
325 J O H
K F
300
275
D
250
300 310 320 330 340
Phase Angle Mix T
(b)
Figure 45. Correlation of Phase Angle method mixing temperature
with the pugmill mixer temperature for 98% coating: (a) all data;
(b) excludes Binder E.
Table 36. Summary of correlation statistics.
Steady Shear Flow Phase Angle
Residual Residual
Mix Test R2 P-value R2 P-value
Error Error
Bucket 22.1 40.8 0.025 25.3 22.4 0.121
Pugmill 27.3 34.6 0.044 25.7 41.9 0.023
Workability 34.0 30.6 0.050 30.5 44.3 0.013
Compaction 18.7 68.4 0.002 16.7 74.8 0.001
Producers' Midpoint 7.8 70.1 0.000 8.8 58.2 0.004
OCR for page 64
Prod. Midpoint = 136.0 + 0.5724 SSF Mix T
340 Regression
N G 95% CI
S 7.79965
I
330 R-Sq 70.1%
R-Sq(adj) 67.4%
O H
320
Prod. Midpoint
F
J C B
310
D E
300
K M
290
280
280 290 300 310 320 330 340
SSF Mix T
Figure 46. Correlation of the SSF mixing temperature with the
midpoint of the binder producers' recommended mixing range.
Prod. Midpoint = 170.9 + 0.4468 Phase Angle Mix T
340 Regression
G
N 95% CI
S 12.1468
I
330 R-Sq 27.4%
R-Sq(adj) 20.8%
H
Prod. Midpoint
O
320
F
B C
J
310
D
E
300
K M
290
300 310 320 330 340
Phase Angle Mix T
(a)
Prod. Midpoint = 115.7 + 0.6269 Phase Angle Mix T_1
Regression
340
95% CI
N
G
S 8.81614
I R-Sq 58.2%
330 R-Sq(adj) 54.0%
Prod. Midpoint
O H
320
F
J B C
310
D
E
300
K
290
300 310 320 330 340
Phase Angle Mix T_1
(b)
Figure 47. Correlation of the Phase Angle Method mixing
temperature with the midpoint of the binder producers'
recommended mixing range: (a) all data; (b) excludes
Binder M.
OCR for page 65
65
EWT = - 27.3 + 1.128 SSF mid
Regression
375 95% CI
I
H S 34.0257
C R-Sq 30.6%
N
350 R-Sq(adj) 24.3%
M
325
EWT
D B
300 E F
G
275
K J
O
250
270 280 290 300 310 320 330
SSF mid
Figure 48. Correlation of SSF mixing and compaction temperature midpoint
with workability experiment equivalent torque temperature.
recommended mixing temperatures. These statistics are based method statistics appeared to be slightly better than the other
on the correlations without suspected outliers, even where correlations with mix test results.
the statistics did not improve with these data excluded. Lower Statistical analyses were conducted to determine whether
P-values indicate the correlation is more significant. A P-value the SSF method or the Phase Angle method provided a better
of 0.05 or less is generally considered a statistically significant overall fit to the experimental data from the mixture tests and
correlation. In general, the correlation statistics are similar for the binder producers' recommended mixing temperatures.
the two methods. The SSF method had better correlation sta- The analyses used an F-statistic to test the null hypothesis that
tistics with the bucket mixer coating test results and the pro- the residual variances were equal, H0: 2(SSF) = 2(PA), for
ducers' recommended mixing temperatures. The Phase Angle each regression summarized in Table 36. The F-test is an
EWT = - 241.3 + 1.785 Phase Angle Mid
Regression
375
95% CI
I
H
C S 30.4978
N R-Sq 44.3%
350
R-Sq(adj) 39.2%
M
325
B
EWT
D
300 F E
G
275
KJ
O
250
290 300 310 320 330
Phase Angle Mid
Figure 49. Correlation of Phase Angle method mixing and compaction
temperature midpoint with workability experiment equivalent torque
temperature.
