The Prediction of Coarse Aggregate Performance by Micro-Deval and Other Soundness, Strength, And Intrinsic Particle Property Tests

7/28/2019 The Prediction of Coarse Aggregate Performance by Micro-Deval and Other Soundness, Strength, And Intrinsic Part

1/30

Lane, Range, Fowler, and Allen 1

THE PREDICTION OF COARSE AGGREGATE PERFORMANCEBY MICRO-DEVAL AND OTHER SOUNDNESS, STRENGTH, ANDINTRINSIC PARTICLE PROPERTY TESTS

ALEXANDER P. LANGThe University of Texas at Austin

PETER H. RANGEThe University of Texas at Austin

DAVID W. FOWLERThe University of Texas at Austin

JOHN J. ALLENThe University of Texas at Austin

ABSTRACT

This research project concentrated on determining whether or not a correlation

existed between laboratory aggregate tests and observed aggregate field performance. For

this purposes, aggregate samples were collected from the majority of the U.S. states as

well as several Canadian provinces and subjected to a variety of strength, soundness, and

intrinsic particle property tests. Additionally, performance data on the aggregates was

obtained for hot-mix asphalt, portland cement concrete, base course, and open-graded

friction course. Numerical and qualitative analyses were performed to evaluate the

success of separating good performers from fair and poor performers using the micro-

Deval test alone as well as the micro-Deval test combined with another test. Furthermore,

attempts were made to determine if a correlation exists between any two tests.

Keywords: micro-Deval; hot-mix asphalt, portland cement concrete, base course, open-

graded friction course; field performance; durability; soundness; abrasion


2/30


1. INTRODUCTION

The International Center for Aggregate Research (ICAR) initiated ICAR Project

507 to evaluate the effectiveness of micro-Deval in predicting aggregate field

performance. Furthermore, the project strived to determine if better performance

prediction was possible using micro-Deval in combination with another test. To

accomplish these tasks, a suite of strength and soundness tests was carried out. The

primary goal was to study micro-Deval, both alone and in combination with other tests.

Finally, correlations were studied to determine how well one test results can be used to

predict another tests results, if at all possible.

To ensure national acceptance of the research results, a comprehensive suite of

aggregates was necessary. To accomplish this task, all departments of transportation

within the United States as well as all Canadian provinces were contacted to participate in

the study by providing aggregates and their observed field performance data. The vast

majority responded resulting in 117 sources representing various mineralogical and

geological spectra. More importantly, however, attempts were made to ensure sufficient

number of good-performing aggregates as well as poor-performing aggregates to produce

meaningful results.

Upon reception of aggregates at the Pickle Research Center at the University of

Texas at Austin, the following tests were performed on all sources according to current

ASTM and AASHTO specifications: micro-Deval, magnesium sulfate soundness,

Canadian freeze-thaw soundness, L.A. abrasion, aggregate crushing value, aggregate

crushing value (saturated, surface-dry), absorption, specific gravity (bulk, saturated

surface-dry, apparent), particle shape factor determination, and percent fractured.


3/30


Appropriate departments of transportation (DOTs) were then contacted to obtain

field performance information within the following major uses: hot-mix asphalt, Portland

cement concrete, base course, and open-graded friction course. Qualitative and

quantitative analyses were then performed to evaluate how well micro-Deval as well as

various two-test combinations can predict field performance within each application.

Within each aggregate usage category, aggregates were divided by their mineralogical

types in order to conclude whether or not additional information could be gained. This

report addresses only hot-mix asphalt and Portland cement applications due to space

constraints. However, the reader is encouraged to refer to the ICAR 507 final report to be

published in August 2006 for the complete analysis. Finally, tests correlations were

studied using all aggregates regardless of aggregate performance in the field.

2. PERFORMANCE CRITERIA

A three-tier rating system was developed for use in this project and is shown in

Table 1. The non-chemically related failures mentioned in the table include, but are not

limited to, such failures as pop-outs, excessive degradation and production of plastic

fines, rutting, D-cracking, and any other failure relating only to the physical properties of

the aggregate.


4/30


This three-tier system provided a way for transportation agencies to easily

identify applicable aggregate while providing the research team a simple but effective

way to objectively rate aggregates. The next step was to attempt to remove bias and

variability from the aggregate performance rating system. Therefore, a standardized

questionnaire was created to be used in discussion with government transportation

agency representatives. Project personnel used this questionnaire to obtain information

concerning each aggregate source such as

- the applications in which the aggregate was used,

- years in service for each application,

- average daily traffic for each application,

- yearly exposure to freeze-thaw,

- characteristic of failures if applicable, and

- time until failure, if applicable.

-

The information obtained by the use of this survey for each source was compared

to the field performance criteria to determine the final field performance rating. Only

Evaluation Description

GoodUsed for 10 or more years with no reported non-chemically relatedfailures

FairUsed at least once where minor non-chemically related failuresrequire repair, but life extends beyond 10 years.

PoorUsed at least once with severe degradation or failure occurringwithin 2 years of service or during construction which severelyinhibits or prevents the use of the application

Table 1: Performance Criteria Developed for Use in This Project


5/30


6/30


freezing and thawing tests (Volger and Grove, 1989). The Canadian Standards

Association (CSA) specification CSA A23.2-24A was chosen for this research.

To determine aggregate strength, the aggregate crushing value test was used with

both oven dry and saturated surface dry conditions. Due to limited use of the aggregate

crushing value test in the U.S., the British Standard, BS 812: Part 110, was chosen

(British Standards Institution, 1990). The relationship between pore characteristics and

aggregate size with durability is well documented. For this reason, absorption and

specific gravity were tested according to AASHTO T 85 specification (AASHTO, 1999).

To test aggregate shape and texture, the flat and elongated test and percent

fractured particles test were chosen. The ASTM D 4791 test method was used for the flat

and elongated test, and ASTM D 5821 test method was used for the percent fractured

particles test. Furthermore, Aggregate IMagining System (AIMS) analysis was performed

by Dr. Eyad Masad at the Texas A&M University on several sources to obtained

additional information discussed in the final report. It is known that the mineralogy of an

aggregate can tell a lot about the probable test results through comparison with

aggregates of similar mineralogical backgrounds. For the purposes of this project, a

professional petrographer was employed to determine the rock type and major

mineralogy of each aggregate.

4. RESULTS DISCUSSION

4.1METHODOLOGY

The results and correlations section for each application was divided into a

discussion of all of the applicable data points for that application, followed by individual

discussions of the applicable rock type subgroups for that application. Only rock type


7/30


subgroups which contained enough data points to provide meaningful correlations were

presented.

With each figure, a table was provided, which detailed the percentage of good,

fair, and poor performers which are correctly categorized. Through visual observation of

pervasive trends, it was determined to divide the performance ratings into two groups;

one being the good performers, the other the fair and poor performers. For the individual

test results, if the good performers fell below the various trial percent losses, they were

considered successfully categorized and if the fair and poor performers were above the

losses, they were considered successfully categorized. The specific gravity results worked

in reverse as the better performers had higher specific gravity values. For the test

combinations, usually a quadrant was drawn which represented the loss limits of the two

tests. If the good performers fell within the quadrant, they were considered correctly

categorized, and if the fair and poor performers fell outside the quadrant, they were

considered correctly categorized. The loss limit or limits which presented the best overall

percentage of correctly categorized points was referred to as the optimized solution.

4.2HOT-MIX ASPHALT

Hot-mix asphalt category provided an excellent division of aggregates by

performance 52 good performers, 26 fair performers, and 20 poor performers. Table 2.1

provides the summary of the statistical analysis performed for hot-mix asphalt aggregates

during this research project. The first column represented the success rates obtained when

only one test was used to separate the good-performing aggregates while the second

column provided the success rates of two-test combinations involving micro-Deval.


8/30


TEST ALONE MICRO-DEVAL COMBINATION

MICRO-DEVAL 69% NOT APPLICABLEMAGNESIUM SULFATE SOUNDNESS 64% 70%L.A. ABRASION 61% 69%CANADIAN FREEZE-THAW 65% 73%

AGGREGATE CRUSHING VALUE 59% 69%AGGREGATE CRUSHING VALUE (SSD) 58% 69%ABSORPTION 58% 70%SPECIFIC GRAVITY (BULK) 54% 68%PARTICLE SHAPE FACTOR 58% 70%PERCENT CRUSHED (1+) 54% NOT APPLICABLEPERCENT CRUSHED (2+) 54% NOT APPLICABLE

Table 4.1: Success Rates Summary

Figure 4.1 illustrated the performance spread of the aggregates for the micro-

Deval test alone. To test how well micro-Deval could separate good performers from fair

and poor aggregates, a micro-Deval threshold value was varied in MS Excel, with correct

percentage predictions computed for each trial value for good, fair, and poor aggregates.

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

Performance

Micro-Deval,%L

oss

Figure 4.1: Micro-Deval vs. Performance

Poor Fair Good


9/30


A weighted percentage of the three was computed as well to see how good the

prediction was overall. Table 4.2 provided the key values to demonstrate how the overall

percentage first increased and then decreased as the micro-Deval threshold value was

increased. Thus, it can be concluded that micro-Deval alone could separate the good

aggregates from the fair and poor sources with the maximum overall success rate of 69%.

Using the same procedure, overall success rates were computed for single tests as shown

in Table 4.1.

TRIAL NUMBER MD Value GOOD FAIR POOR OVERALL

1 7% 17% 100% 95% 55%

2 8% 31% 100% 90% 61%

3 11% 54% 73% 90% 66%

4 12% 56% 73% 90% 67%

5 13% 63% 65% 90% 69%

6 14% 67% 54% 85% 67%

7 17% 75% 42% 70% 65%

8 18% 77% 38% 65% 64%

9 20% 81% 27% 40% 58%

10 21% 83% 23% 30% 56%

Table 4.2: Success Rate

The data in the second column indicated that success rates could be improved

from the highest of 69% by using micro-Deval alone to 73% overall success rate by using

micro-Deval and Canadian freeze-thaw tests together. This result was in line with

research performed by Chris Rogers et al (Senior and Rogers, 1991), who concluded that

better classification results were obtained by combining Canadian freeze-thaw test data

with micro-Deval test data. Figure 4.2 displayed the results when micro-Deval was

plotted against Canadian freeze-thaw. The three quadrants outlined in the plot represent

the results of numerous trials in an attempt to optimize overall prediction success

percentage. During the first attempt, outlined by the solid line, only good performers


10/30


were included in the quadrant, resulting in an overall success rate of 63%. By including

several fair sources into the quadrant area, overall success rate of 69% was achieved. By

further extending the border to the right, the overall percentage increased to the

maximum value of 73%. Numerical results for the three trials were summarized in Table

4.3. As the table indicated, the overall success rate is achieved at the expense of

decreasing fair and

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0

Micr o-Deval, % Los s

CanadianFreeze-Thaw

,%L

oss

POOR

FAIR

GOOD

Figure 4.2: Canadian Freeze-Thaw vs. Micro-Deval

TRIAL NUMBER 1 2 3

CFT 5 5 5

MD 8 10 19

QUADRANT 100% 89% 75%

GOOD 31% 48% 75%

FAIR 100% 88% 62%POOR 100% 100% 85%

OVERALL 63% 69% 73%



11/30


poor prediction success rates as well as the quadrant prediction success rate. Thus, it

could be concluded that the additional information provided by the Canadian freeze-thaw

test was quite significant as its inclusion in the analysis leads to the higher overall success

rate than that of using the micro-Deval test alone. Furthermore, it was clear from Table

4.1 that several other two-test combinations yielded higher prediction success rates than

using any of the individual tests. Thus, it was the conclusion of this research project team

that micro-Deval represented the best single-test prediction performance while the

combination of micro-Deval and Canadian freeze-thaw tests represented the best two-test

combination for performance classification of aggregates used in the hot-mix asphalt.

4.3HOT-MIX ASPHALT LIMESTONE AND DOLOMITE

Similar to the comprehensive section, the test combination with the best

correlation to field performance for limestone and dolomite was micro-Deval and

Canadian freeze-thaw. Table 4.4 illustrated the combination of these two tests increased

the success rate from 71% for micro-Deval alone to 74%. From Table 4.5, the bounds

specified for this combination only successfully qualified 53% of the good performers; so

they would not be good for eliminating potential sources. However, if a source fits within

these bounds it will most likely perform well.

Test Test Alone Micro-Deval Combination

Micro-Deval 71% N/A

Magnesium Sulfate Soundness 65% 71%

L.A. Abrasion 61% 71%

Canadian Freeze-Thaw 61% 74%

Aggregate Crushing Value 68% 71%Aggregate Crushing Value (SSD) 57% 71%

Absorption 61% 71%

Specific Gravity (Bulk) 58% 71%

Particle Shape Factor 71% 71%

Table 4.4: Success Rate Summary


12/30


Micro-Deval in combination with Canadian Freeze-Thaw provided the most

meaningful correlations. From Figure 4.3 and Table 4.5, the bounds from the

comprehensive section (refer to trial 3) represent good limits for good performers. The

best overall success rate was represented by the blue lines which successfully categorized

74% of the total.

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0

Micro-Deval, % Loss

CanadianFreeze-Thaw,%Loss

POOR

FAIR

GOOD


Trial 1 2 3 4

CFT 5% 5% 5% 6%

MD 8% 10% 19% 18.6%

Quadrant N/A 67% 78% 89%

Good N/A 13% 47% 53%

Fair N/A 88% 88% 88%

Poor N/A 100% 88% 100%

Overall N/A 55% 68% 74%



13/30


4.4HOT-MIX ASPHALT SILICEOUS GRAVEL

For the siliceous gravel subgroup of hot-mixed asphalt, several combinations of

tests provided good correlations with field performance. As Table 4.6 indicated, the best

combinations were micro-Deval with Canadian freeze-thaw, aggregate crushing value

(SSD), and particle shape factor as well as Canadian freeze-thaw with aggregate crushing

value. The small amount of data points compared to the previous section increased the

relative weight each data point carried. Based on this, it was difficult to draw

conclusions, but it appeared as though the performance of siliceous gravels could be more

accurately predicted with lab tests than most aggregate types.

Micro-Deval in combination with Canadian Freeze-Thaw provided one of the

more meaningful correlations for the siliceous gravel subgroup. From Figure 4.4 and

Table 4.7, the bounds from the comprehensive section were effective in isolating the

good performers (refer to trials 1, 2, and 3). The best possible solution was detailed in

trial 4 and illustrated with the blue line. These limits produced an overall success rate of

93%. The lone fair point within these bounds, designated WA-1, had mixed performance


Micro-Deval 79% N/A




Aggregate Crushing Value 79% 86%

Aggregate Crushing Value (SSD) 79% 93%

Absorption 57% 79%





14/30


results. In dry environments it performed well, but in wet environments it performed

poorly.

0.0

2.0

4.0

6.0

8.0

10.0

12.0

0.0 5.0 10.0 15.0 20.0 25.0 30.0

Micro-Deval, % Loss


POOR

FAIR

GOOD


Trial 1 2 3 4

CFT 5% 5% 5% 4%

MD 8% 10% 19% 11%

Quadrant 100% 80% 71% 83%

Good 40% 80% 100% 100%

Fair 100% 80% 80% 80%

Poor 100% 100% 75% 100%

Overall 79% 86% 86% 93%



15/30


As is illustrated in Figure 4.5 and Table 4.8, the results for aggregate crushing

value (SSD) and micro-Deval provided very good correlations with performance. The

limits described in trial 4 and illustrated by the blue line successfully categorized 93% of

the aggregates - an increase of 14% over micro-Deval alone.

0

5

10

15

20

25

30

35

0.0 5.0 10.0 15.0 20.0 25.0 30.0

Micro-Deval, % Loss

AggregateCrush

ingValue(SSD),%Loss

POOR

FAIR

GOOD

Figure 4.5: Aggregate Crushing Value (SSD) vs. Micro-Deval

Trial 1 2 3 4

WCV 22% 33% 33% 17%

MD 5% 8.5% 13.5% 10%

Quadrant 100% 67% 63% 100%

Good 20% 40% 100% 80%

Fair 100% 100% 60% 100%

Poor 100% 75% 75% 100%

Overall 71% 71% 79% 93%



16/30


The combination of particle shape factor and micro-Deval proved surprisingly

effective for this subgroup. From Figure 4.6 and Table 4.9, the optimized solution

resulted in an overall success rate of 93%.

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

4.50

0.0 5.0 10.0 15.0 20.0 25.0 30.0

Micro-Deval, % Loss

Particle

ShapeFactor

POOR

FAIR

GOOD

Figure 4.6: Particle Shape Factor vs. Micro-Deval

Trial 1 2 3 4

PSF 2.5 2.6 4 2

MD 5% 7.5% 12.5% 11%

Quadrant 100% 100% 63% 100%

Good 20% 40% 100% 80%

Fair 100% 100% 60% 100%

Poor 100% 100% 75% 100%

Overall 71% 79% 79% 93%



17/30


From Figure 4.7 and Table 4.10, the combination of aggregate crushing value and

Canadian freeze-thaw resulted in high overall success rate of 93%. From the figure, it

was clear Canadian freeze-thaw was a very effective test at eliminating poor performers

within this subgroup.

0

5

10

15

20

25

30

0.0 2.0 4.0 6.0 8.0 10.0 12.0

Canadian Freeze-Thaw, % Loss

AggregateCrushingValue,%Loss

POOR

FAIR

GOOD

Figure 4.7: Aggregate Crushing Value vs. Canadian Freeze-Thaw

Trial 1 2 3 4

ACV 25% 28% 28% 22%

CFT 1.9% 1.9% 4.4% 4%

Quadrant 100% 100% 71% 83%

Good 20% 20% 100% 100%

Fair 100% 100% 60% 80%

Poor 100% 100% 100% 100%Overall 71% 71% 86% 93%



18/30


4.5HOT-MIX ASPHALT GRANITE

The granite subgroup of hot-mix asphalt contained only fair and good performers.

The most effective correlation was provided by the micro-Deval test alone. As Table 4.11

illustrated, no combination of tests was able to improve the success rate above 92%. One

good performing aggregate, QC-1, could not be isolated from the fair performers by any

combination of tests.

From Figure 4.8 and Table 4.12, a clear delineation existed between the two

categories at 8% micro-Deval loss. Only one data point, a good performer designated

QC-1, fell on the wrong side of the line resulting in an overall success rate of 92%; this

represented a 25% improvement over the total percentage of good performers. None of

the test combinations were able to isolate this one outlying point. Thus, the overall

success rate did not improve to above 92%.


Micro-Deval 92% N/A




Aggregate Crushing Value 75% 92%

Aggregate Crushing Value (SSD) 75% 92%

Absorption 42% 92%





19/30


0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

Performance

Micro-Deval,%Loss

Figure 4.8: Micro-Deval vs. Performance

Trial MD Good Fair Overall

1 7% 38% 100% 58%

2 8% 88% 100% 92%

3 11% 100% 25% 75%

4 12% 100% 25% 75%5 13% 100% 0% 67%

6 14% 100% 0% 67%

7 17% 100% 0% 67%

8 18% 100% 0% 67%

9 20% 100% 0% 67%

10 21% 100% 0% 67%


Poor Fair Good


20/30


4.6PORTLAND CEMENT CONCRETE

From Table 4.13, combinations of tests did not improve the overall success rate

considerably for predicting field performance in Portland cement concrete. However, the

success rates for the quadrants were quite good when micro-Deval was combined with

magnesium sulfate soundness, Canadian freeze-thaw, absorption, or specific gravity. The

micro-Deval and Canadian freeze-thaw combination produced the best results.


Micro-Deval 83% N/A

Magnesium Sulfate Soundness 81% 85%Canadian Freeze-Thaw 88% 88%

Absorption 83% 83%



The correlation between field performance and the combination of micro-Deval

and Canadian freeze-thaw test results was illustrated in Figure 4.9. As detailed by the

solid lines and trial 1 of Table 4.14, the majority of the good performers could be

bounded by a Canadian freeze-thaw loss of less than 3.6% and a micro-Deval loss of less

than 21% with no fair or poor performers. These bounds successfully qualified 77% of

the good performers.


21/30


0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0

Micro-Deval, % Loss

CanadianFreeze-Thaw,%Lo

ss

POOR

FAIR

GOOD


Trial 1 2

CFT 3.6% 6.5%

MD 21% 21%

Quadrant 100% 90%

Good 77% 95%

Fair 100% 50%

Poor 100% 60%

Overall 81% 88%



22/30


4.7PORTLAND CEMENT CONCRETE LIMESTONE AND DOLOMITE

The results for limestone and dolomite showed how much more difficult it is to

predict performance compared to the comprehensive section. From Table 4.15, the

success rate could not be improved for any combination of tests over one test alone. Such

behavior is not entirely unexpected as the material properties of limestones and dolomites

could vary quite drastically.

4.8PORTLAND CEMENT CONCRETE SILICEOUS GRAVEL

The results for the siliceous gravel sub-group showed combinations of Canadian

freeze-thaw, absorption, or specific gravity with micro-Deval were strong predictors of

field performance in Portland cement concrete. As summarized in Table 4.16, all three of

these combinations resulted in perfect performance prediction.


Micro-Deval 89% N/A



Absorption 78% 100%




Micro-Deval 75% N/A


Canadian Freeze-Thaw 83% 83%Absorption 58% 75%




23/30


As was shown in Figure 4.10 and Table 4.17, all of the good performers in the

siliceous gravel subgroup fit within the overall bounds of 3.6% Canadian freeze-thaw

loss and 21% micro-Deval loss as defined in trial 1 for the comprehensive section. These

bounds successfully qualified 100% of all categories.

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

0.0 5.0 10.0 15.0 20.0 25.0

Micro-Deval, % Loss


POOR

FAIR

GOOD

Figure 4.10: Canadian Freeze-Thaw vs.

Trial 1 2

CFT 3.6% 6.5%

MD 21% 21%

Quadrant 100% 75%

Good 100% 100%

Fair 100% 0%

Poor 100% 50%

Overall 100% 78%



24/30


Micro-Deval

From Figure 4.11 and Table 4.18, a micro-Deval loss less than 11% and

absorption of less than 2% isolated all of the good performers (refer to blue lines and trial

5). These bounds successfully qualified 100% of all categories.

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

0.0 5.0 10.0 15.0 20.0 25.0

Micro-Deval, % Loss

Absorpti

on,%

POOR

FAIR

GOOD

Figure 4.11: Absorption vs. Micro-Deval

Trial 1 2 3 4 5

ABS 1.2% 1.2% 2.0% 2.0% 2.0%

MD 18% 21% 18% 21% 11%

Quadrant 100% 100% 86% 86% 100%

Good 33% 33% 100% 100% 100%

Fair 100% 100% 100% 100% 100%

Poor 100% 100% 50% 50% 100%

Overall 56% 56% 89% 89% 100%



25/30


Specific gravity also had very good correlation with field performance when

combined with micro-Deval. From Figure 4.12 and Table 4.19, the limits of specific

gravity greater than 2.5 and micro-Deval loss less than 11%, successfully qualified all of

the aggregates (refer to blue lines and trial 2).

2.20

2.30

2.40

2.50

2.60

2.70

2.80

2.90

3.00

0.0 5.0 10.0 15.0 20.0 25.0

Micro-Deval, % Loss

SpecificG

ravity(Bulk)

POOR

FAIR

GOOD

Figure 4.12: Specific Gravity (Bulk) vs. Micro-Deval

Trial 1 2

SG(Bulk) 2.45 2.50

MD 21% 11%

Quadrant 90% 100%

Good 95% 100%

Fair 75% 100%

Poor 40% 100%

Overall 87% 100%



26/30


4.9TEST CORRELATIONS

One of the objectives of this research project was to determine if any correlations

could be drawn between any two tests conducted in the study. To accomplish this task,

test results of one test were plotted against test results of another test, regardless of

performance rating. Once the graphs were created, correlation analysis was performed on

each one to measure the strength of the association between numerical variables by

performing linear regression analysis. In cases where no linear relationship clearly

existed, logarithmic, polynomial, power, and exponential regression analyses were

performed using Microsoft Excel 2003. During the analysis, the coefficient of

determination R2, which measures the proportion of variation that is explained by the

independent variable X in the regression model, was computed according to the

following formula:

squaresofsumtotal

squaresofsumregression

SST

SSRr

___

___2==

where SST measures the variation of the Y-values around their mean Y and SSR explains

the variation attributable to the relationship between X and Y. Finally, the correlation

coefficient could be computed by taking the square root of the R2

value. Since it is

common in research practice to report the coefficient of determination value, R2

will be

used throughout this report for comparison and analysis purposes.

Upon recommendation of Dr. Zhanmin Zhang of the University of Texas at

Austin, whose expertise lies in the area of statistical analysis within the field of Civil

Engineering, R2

values were computed for three cases:


27/30


1. The complete data set with no outliers eliminated.

2. The data set with 99.7% of the values retained and outliers eliminated

based on the assumption of normal population distribution and using the

interval of *3 , where is the sample mean and is the sample

standard deviation.

3. The data set with 95% of the values retained and outliers eliminated based

on the assumption of normal population distribution and using the interval

of *2 , where once again is the sample mean and is the sample

standard deviation.

The complete data set analysis was discussed in this report while the other two

cases were included in the full ICAR 507 report to be published in August 2006 since

data reductions did not produce significant improvements in the results. Table 4.20

provided the R2

results for the complete data set.

MD

MSS

LAA

CFT

ACV

ACV(SSD)

ABS

SG(

BULK)

SG(SS

D)

SG(AP

P)

MSS 0.536

LAA 0.116 0.179

CFT 0.320 0.389 0.041

ACV 0.220 0.124 0.650 0.012

ACV(SSD)

0.196 0.080 0.487 0.005 0.836

ABS 0.401 0.536 0.075 0.147 0.141 0.134

SG(BULK)

0.172 0.311 0.171 0.031 0.167 0.172 0.647

SG (SSD) 0.114 0.247 0.180 0.020 0.161 0.169 0.492 0.975

SG (APP) 0.012 0.079 0.168 0.009 0.122 0.133 0.140 0.744 0.854

PSF 0.037 0.013 0.000 0.004 0.033 0.028 0.004 0.009 0.011 0.011

Table 4.20: Complete Data Set R2 Values


28/30


5.SUMMARY AND CONCLUSIONS

Throughout all of the test results for all of the various applications and subgroups,

the tests which most consistently provided correlations with field performance, either

alone or in combination with other tests, were micro-Deval, Canadian freeze-thaw,

absorption, and specific gravity. Several correlations indicated specific loss limits which

seemed to correctly isolate good performers from the rest. These limits would best be

used to positively identify sources which would likely perform well in the field. They

should not be used to eliminate aggregates as in most cases a large percentage of good

performers were outside the bounds as well.

Additionally, correlation analysis yielded several notable results. Firstly, micro-

Deval did not have statistically significant correlations to any other tests. However, the

following pairs of tests were found to have significant correlations: L.A. abrasion and

aggregate crushing value, L.A. abrasion and aggregate crushing value (saturated, surface-

dry), aggregate crushing value and aggregate crushing value (saturated, surface-dry),

absorption and bulk specific gravity, bulk specific gravity and saturated surface-dry

specific gravity, bulk specific gravity and apparent specific gravity, as well as saturated

surface-dry specific gravity and apparent specific gravity. Interestingly, magnesium

sulfate soundness test did not correlate well to Canadian freeze-thaw test. Finally, particle

shape factor test had extremely low R2 values and hence did not correlate at all to any of

the tests carried out during this research.


29/30


REFERENCES

American Association of State Highway and Transportation Officials, "Standard TestMethod for Soundness of Aggregate by Use of Sodium Sulfate or Magnesium Sulfate," T104-99, 1999.

American Association of State Highway and Transportation Officials, "Standard TestMethod for Resistance of Coarse Aggregate to Degradation by Abrasion in the Micro-Deval Apparatus," TP58-00, 1999.

American Association of State Highway and Transportation Officials, "Standard Methodof Test for Soundness of Aggregates by Freezing and Thawing," T 103-91, 2000.

British Standards Institution, "Methods for Determination of aggregate crushing value(ACV)," BS 812: Part 110, 1990.

Hanna, A., K. Folliard, and K. Smith, "Aggregate Tests for Portland Cement ConcretePavements: Review and Recommendations,"Research Results Digest, No. 281, NCHRP,2003.

Meininger, R. C., "Degradation Resistance, Strength, and Related Properties,"Significance of Tests and Properties of Concrete and Concrete Making Materials, ASTMSpecial Publication, STP 169C, American Society for Testing and Materials, 1994, p.388.

Senior, S. A., and C. A. Rogers, "Laboratory Tests for Predicting Coarse AggregatePerformance in Ontario," Transportation Research Record, No. 1301, 1991, pp. 97-106.

Volger, R., and G. H. Grove, "Freeze-Thaw Testing of Coarse Aggregate in Concrete:Procedures Used by Michigan Department of Transportation and Other Agencies,"Cement, Concrete and Aggregates, Vol. 11, No. 1, American Society for Testing andMaterials, 1989, pp. 57-66.


30/30


The Prediction of Coarse Aggregate Performance by Micro-Deval and Other Soundness, Strength, And Intrinsic Particle Property Tests

Documents