Report

Estimating Dry Density

From
Soil Stiffness & Moisture Content

29 June, 1999

Humboldt Mfg. Co.

7300 West Agatite
Norridge, IL 60656-4704

Scott Fiedler, Product Manager Melvin Main, Marketing

Humboldt Mfg. Co.
7300 West Agatite Ave., Norridge, IL 60656, U.S.A.
800-544-7220 extension 231 (Voice)
708-456-0137 (Fax)
fiedler_hmc@msn.com (Email)
Main Associates
16 Vegas Dr., Hanover, PA 17331, U,S,A.
717-637-8246 (Voice & Fax)
main@cyberia.com (Email)
Problem
If any new method of evaluating soil compaction is to be widely accepted, a firm
relationship must be established between this method and the most accepted current
methods, measurements of dry density.

Objective
Develop an analytical-empirical relationship between soil stiffness and density. Validate
the relationship with data from Humboldt GeoGauge™ measurements and accepted
methods of measuring density.

Approach
Began with the analytical-empirical relationship that was developed by BBN
Technologies of Cambridge, MA some 4 years ago from the work of Hryciw & Thomann
1
.
ρ0
ρD = C .5
1)
1 + 1.2 [ - .3].5
K
where
(C1 σ1P)4a
C=
(1-υ)
C1 = is a function of moisture and soil type
σ = is the overburden stress
1

P = is typically between 1/2 and 1/4

a = is the foot radius
υ = is Poisson’s ratio
ρ D = is the dry density
ρ = is the ideal, void free density
0

K = is stiffness

Define C for a geographical region or group of soil classes, independent of everything but
moisture. Do this based on companion stiffness, moisture content and density
measurements. Then use C, measured stiffness and measured moisture content to
estimate dry density. Compare the estimates to density measurements made with a
nuclear gauge and sand cone.

1
Roman D. Hryciw & Thomas G. Thomann, “Stress-History-Based Model for Cohesionless Soils”,
Journal of Geotechnical Engineering, Vol. 119, No, 7, July, 1993

1
Results

Analytical-Empirical Relationship
Early attempts at following this approach revealed two things.
• A more precise estimation was possible when moisture content was broken out of
the constant C and
• More precision was possible when the values of C were calculated from a linear
relationship with stiffness and moisture content.

Solving equation 1) for C yields

C=K {[(ρ /ρ -1)/1.2] + 0.3}

0 D
2
2)

If we let C = Cm, where m = (% moisture content by weight)/100), then C can be

represented as
2
C = (K/m) {[(ρ /ρ -1)/1.2] + 0.3}
0 D
3)

This representation allows for moisture content to be included in each estimate of dry
density. It also allows the values of C determined from the companion measurements to
be fitted to a linear equation with our two independent variables, K and m.

C = n(K/m.25) + b 4)

where
n is the slope
and
b is the intercept.

This linear relationship between C, K and m allows a more appropriate value of C to used
in the estimate of each dry density as opposed to selecting a limited number of Cs to used
over several moisture ranges. Breaking m out of C and using this linear relationship
provided closer agreement between measured and estimated dry density in 23% of the
cases compared to not doing either.

Numerous other modifications of equation 1) were numerically analyzed relative to

actual companion measurement data. The analytical-empirical relationship represented
by equations 3 and 4 fit the data the best. Figure 1 is a 3D surface plot of K, M and ρD as
described by this relationship. The relationship appears to be well behaved in the ranges
of density, stiffness and moisture content that most applications will encounter.

Based on the usage of current methods for evaluating compaction and a consensus of
GeoGauge™ customers, the following criteria were established for the evaluation of the
above approach.

2
 a) Estimates of dry density should be within 5% of the measured values about 70% of
the time & within 10% > 90% of the time.
b) The span of measured & estimated densities should be almost the same.
c) A one-to-one correspondence of measured to estimated densities should yield a
correlation coefficient of > .3 (typically > .5).

Validation of the Relationship

Five hundred and seventy seven (577) companion measurements were made in
California, Ohio, Florida, Missouri, New York, North Carolina and Virginia by the
FHWA, California Polytechnic Institute, the H. C. Nutting Co., the City of San Jose, the
FDOT, the MODOT, the NYSDOT and the NCDOT. These measurements were made
largely independent of Humboldt. The data, the estimates of dry density and the
comparisons of the estimates to direct measurements are presented in Appendices 1
through 9.
Each appendix contains the following information.
• Multiple plots of raw data; density vs. stiffness vs. moisture content
• Summaries of how well C was determined form a function of stiffness & moisture
segregated by groups of similarly performing soils
• Plots of estimated vs. measured density in terms of percentage difference and one-
to-one correspondence
• All the data used to determine C and numerical data for all density estimates,
segregated by data that was used to determine C and data that was not2

It was evident that several classes or groups of similarly performing soils were
represented by the data from each source. In some cases, when C was plotted against a
function of K and m, the presence of more than one linear relationship was apparent. In
other cases, there was a clustering of values of C that were calculated from companion
measurements. When one or both of these conditions coincided with test sites or
locations, the data was correspondingly segregated and analyzed independently. This
greatly improved the results of the analysis in meeting the criteria stated earlier. Since
only the California Polytechnic Institute provided soil classifications with its data, the
validity of this operation will need to be confirmed with the sources of the data.

It is also evident that the relationship represented by equations 3 and 4 will not provide
satisfactory estimates of density for every soil. Soils due to stabilization additives,
construction methods, site conditions or just their nature are apparently atypical. The
data from the FDOT is a good example. As can be seen from the raw data in Appendices
5 and 6, that sandy, limestone stabilized soil are not typical of the soil behavior illustrated
in the other appendices. For such soils, it was found that by using the relationship
represented by equations 1 and 4 satisfactory estimates of density were possible. Figure 2
is a 3D surface plot of K, M and ρD as described by equations 1 and 4.

2
Due to the volume of data, this information is omitted from the pdf version of the report. A hard copy of
this information is available upon request.

3
Table 1 summaries the results presented in the appendices. The 3 evaluation criteria
applied across the 10 data sources are met 96% of the time. Only criteria a) is missed in
the MODOT data.

Conclusions
An analytical-empirical relationship has been developed that allows the estimation of dry
density from soil stiffness and moisture content within tolerances that are typical in the
use current field measurements. The successful application of this relationship requires
that it be adjusted for groups of similarly performing soils and atypical soils. This
relationship firmly connects soil stiffness, as measured by the Humboldt GeoGauge™,
with dry density. This relationship in conjunction with companion measurements of
moisture content and stiffness is a potential alternative method for determining dry
density.

4
 Figure 1:
Su rface P lo t o f K, M an d ρD
as Described by Equations 3) & 4)

ρ0
ρD = mC .5
1 + 1.2 [ K - .3]

C = n(K/m.25) + b

5
 Figure 2:
Su rface P lo t o f K, M an d ρD
as Described by Equations 1) & 4)

ρ0
ρD = C .5
1 + 1.2 [ K - .3]

C = n(K/m.25) + b

6
 Table 1: Summary of Results

∆%, ρD (GeoGauge) re ρD (Nuc) R2

Data Number of Relationship (percentage of estimates within 5, 10 Density Span (correlation coefficient)
Source Companion Used and 15 % of the direct measurements) GeoGauge/Nuc ρD (Nuc) vs.
Measurements 5% 10% 15% (pcf)
ρD (GeoGauge)
Cal. Poly. 80 Eq. 3 & 4 82% 100% - 32/35 0.83
H.C. 66 Eq. 3 & 4 95% 5% - 34/33 0.86
Nutting
San Jose 120 Eq. 3 & 4 70% 99% 100% 33/27 0.33
FDOT 112 Eq. 1 & 4 88% 100% - 23/18 0.43
(field)
FDOT 34 Eq. 1 & 4 97% 100% - 10/9 0.39
(lab)
MODOT 30 Eq. 3 & 4 60% 100% - 39/36 0.77
NYSDOT 50 Eq. 3 & 4 90% 100% - 0.31/0.34 0.51
Mg/m3
NCDOT 17 Eq. 3 & 4 100% - - 17/16 0.90
FHWA 60 Eq. 3 & 4 88% 100% - 66/62 0.94

7
 Appendix 1
Analysis of NCDOT Data

8
NCDOT Raw Data

9
 NCDOT
Data Analysis Summary

ρD (HSG) re ρre
∆%,∆ρ%:D (GeoGauge) ρD (Nuc)
D (Nuc)

17

15
ρ0
13 ρD = mC .5
1 + 1.2 [ K - .3]
11
C = n(K/m.25) + b

7
D et e r m i n a ti o n o f C

Soil Group #1:

5 C = 3.7095(K/m.25) + 8.2414
R2 = 0.8987
Soil Group #2:
3 C = 5.6425(K/m.25) + 3.4313
R2 = 0.8331
1

-4.0 -2.0 0.0 2.0 4.0 6.0

%. Percent
Measured
Measuredvs.
vs. Predicted
Estimated Density
Density
(NCDOT Data)
95.0
y = 0.8154x
2 + 15.339
R = 0.9029
90.0

85.0

Data
Linear Fit
80.0

75.0
75.0 80.0 85.0 90.0 95.0
ρD (GeoGauge),
ρ D , (HSG), pcf
pcf

10
 Appendix 2
Analysis of Cal. Poly. Data

11
Cal. Poly. Raw Data

12
California Polytechnic Institute, San Luis Obispo, CA
Data Analysis Summary

∆%, ρ∆D%:(GeoGauge)
ρD (HSG) re re
ρ ρ(Nuc)
D
D (Nuc)

76

71
ρ0
66
ρD = mC .5
61 1 + 1.2 [ K - .3]
56
C = n(K/m.25) + b
51

46
D et e r m i n a ti on of C
41
Site #1, AASHTO A-2-6(U):
36 C = 4.4561(K/m.25) + 12.704
R2 = 0.8943
31
Site #2, AASHTO A-6(6):
26 C = 3.765(K/m.25) + 19.165
R2 = 0.8378
21 Site #3, AASHTO A-6(7):
C = 4.2431(K/m.25) – 4.8947
16 R2 = 0.8199
11

1
-10 -5 0 5 10
%. Percent
Measured vs. Predicted Density
Measured vs. Estimated Density
(Cal. Poly. Data)
130
y = 0.9708x
2 + 3.9808
R = 0.8292
120

110
s

100
Data
Linear Fit

90
90 100 110 120 130

ρD (GeoGauge),
ρD (HSG), pcf pcf

13
 Appendix 3
Analysis of H. C. Nutting Data

14
H. C. Nutting Raw Data

15
 H. C. Nutting Co., Cincinnati, OH
Data Analysis Summary

∆%,∆%,
ρD (GeoGauge) re ρ(Nuc)
ρ (HSG) re ρ
D D (Nuc)
D

33

31

29

27

25

23

21
Data Not Used for C D et e r m i n a ti o n o f C
19 Data Used for C Soil Group #1:
17 C = 2.8335(K/m.25) + 11.465
R2 = 0.7417
Soil Group #2:
15
C = 3.1484(K/m. 25) + 2.6727
R2 = 0.9693
13 Soil Group #3:
C = 2.8146(K/m. 25) + 10.44
11 R2 = 0.9414
9

5 Measured vs. Predicted Density

Measured vs. Estimated Density
(H.C. Nutting Data)
3 140

1 Data
130 Linear Fit
-15 -10 -5 0 5 10 15

%, Percent 120

110

100
y = 0.9174x
2
+ 9.489
R = 0.8638
90
90 100 110 120 130 140
ρD (GeoGauge), pcf
ρD
(HSG), pcf

16
 Appendix 4
Analysis of San Jose Data

17
San Jose Raw Data

18
 City of San Jose, CA
Data Analysis Summary

∆%,∆ρ%:D (GeoGauge)
ρ (HSG) re ρre (Nuc)
D D ρD (Nuc)
58
ρ0
55 ρD = mC .5
52 1 + 1.2 [ K - .3]
49 C = n(K/m.25) + b
46
Determination of C
43
200
40
180

37 160

34 Data Not used for C 140

31 Data Used for C 120

100
28
80
25
60

22 40 y = 3.421x + 12.423
R2 = 0.8517
19 20

0
16
0 10 20 30 40 50
0.25
13 K/(m ), MN/m

10
Measured vs, Predicted Density
Measured vs. Estimated Density
(San Jose Data)
7
130
4

1 125
-20 -15 -10 -5 0 5 10 15
120
%, Percent
115

110

105

100

95 y = 0.5763x + 49.73
R2
90 0 3291
90 100 110 120 130
ρD (HSG), pcf
ρD (GeoGauge), pcf

19
 Appendix 5
Analysis of FDOT Field Data

20
FDOT Raw Field Data

21
 FDOT
Field Data Analysis Summary

∆%, ρD (GeoGauge) D ρ ρ(Nuc)

∆%: ρ (HSG) rere D (Nuc)
D

115
ρ0
112
109
ρD = C .5
1 + 1.2 [ K - .3]
106
103 C = n(K/m.25) + b
100
97
D et e rm i n a ti o n o f C
94
91 Soil Group #1:
88 C = 0.3536(K/m.25) + 1.8587
85 R2 = 0.9439
Soil Group #1a:
82
C = 0.4613(K/m.25) + 1.0223
79
R2 = 0.97
76 Soil Group #2:
73 C = 0.5391(K/m. 25 ) + 0.1964
70 R2 = 0.9568
Soil Group #3:
67
C = 0.4126(K/m. 25 ) + 0.8955
64 R2 = 0.9828
61 Soil Group #4:
58 C = 0.1288(K/m.25) + 6.48
55 R2 = 1
52
49 Measured vs, Predicted Density
Measured vs. Estimated Density
(FDOT Data)
46
125
43
40 Data
37 120 Linear Fit

34
31 115
28
25
110
22
19
16 105
13
10
100
7 y = 0.589x
2 + 44.563
4 R = 0.4305

1 95

-10 - 8 -6 -4 -2 0 2 4 6 8
%, Percent ρD (GeoGauge),
ρ (HSG), pcf pcf
D

22
 Appendix 6
Analysis of FDOT Lab Data

23
FDOT Raw Lab Data

24
 FDOT
Lab Data Analysis Summary

∆%:
∆%, ρDρ(GeoGauge)
D
(HSG) re ρ (Lab.)
re ρD (Nuc)
D

17 ρ0
16 ρD = C .5
1 + 1.2 [ K - .3]
15

14 C = n(K/m.25) + b
13
Determination of C
12 14

11 y = 0.3947x + 1.8007
13 2
R = 0.9397
10 12

9 Data Not Used for C

11
Data Used for C
8
10

7
9

6
8

5
7

4
6 Data
Linear Fit
3
5

2
4
0 10 20 30
1 K/m, (MN/M)

- 4 - 3 - 2 - 1 0 1 2 3 4 5
%, Percent Measured
Measured vs.
vs. Estimated Density
Predicted Density
(FDOT Data)
116

y = 0.5861x
2 + 44.97
114 R = 0.3865

112

110

108

Data
106 Linear Fit

104
104 106 108 110 112 114 116
ρD (GeoGauge),
ρ D(HSG), (pcf) pcf

25
 Appendix 7
Analysis of MODOT Data

26
MODOT Raw Data

27
 MODOT
Data Analysis Summary

∆%,∆ρ%:D (GeoGauge) D ρD (Nuc)

ρD (HSG) re ρre (Nuc)

28

25
ρ0
22 ρD = mC .5
1 + 1.2 [ K - .3]
19
C = n(K/m.25) + b
16

13
D et e r m i n a ti o n o f C
10
Soil Group #1:
7 C = 4.5241(K/m.25) – 2.0602
R2 = 0.9239
4 Soil Group #2:
C = 5.9674(K/m.25) – 16.752
1 R2 = 0.9834
-10 -5 0 5 10
%. Percent

Measured
Measuredvs.vs.Predicted Density
Estimated Density
(MODOT Data)
145

y = 0.7527x
2 + 26.806
135
R = 0.7717
125

115

105
Data
95 Linear Fit

85
85 95 105 115 125 135
ρ (HSG), pcf
ρD (GeoGauge),
D
pcf

28
 Appendix 8
Analysis of NYSDOT Data

29
NYSDOT Raw Data

30
 NYSDOT
Data Analysis Summary

∆%,
∆%, ρD ρ(GeoGauge)
(HSG) re ρ(re
N uρc ) (Nuc)
D

ρ0
29 ρD = mC .5
27 1 + 1.2 [ K - .3]

25
C = n(K/m.25) + b
Not Used for C
23 Used for C
Deterimination of C

21 256
Data
19 206
Linear Fit

17
156
15
106
13

11 56 y = 4.1333x
2 + 11.744
R = 0.887
9
6
10 20 30 40 50
7
K/m , MN/m
5
Measured vs. Predisted Density
3
Measured vs. Estimated Density
(NYSDOT Data)

1 2.10 y = 0.7879x + 0.4068

2
2.05 R
-10 -5 0 5 10 0 5073
2.00
%, Percent
1.95

1.90

1.85

1.80 Data
1.75 Linear Fit

1.70
1.70 1.80 1.90 32.00 2.10
ρ D (HSG), Mg/mpcf
ρD (GeoGauge),
31
 Appendix 9
Analysis of FDOT Lab Data

32
FHWA Raw Data

33
 FHWA Turner- Fairbanks
Data Analysis Summary

∆%:
∆%, ρD ρ (HSG) re ρre
(GeoGauge)
D
(Nuc)
D ρ (Nuc)
D

29
ρ0
27 ρD = mC .5
1 + 1.2 [ K - .3]
25 C = n(K/m.25) + b

23
D et e r m i n a ti o n o f C
21
Soil Group #1 (Bridge):
19 C = 7.7675(K/m.25) – 12.337
R2 = 0.5699
Soil Group #2 (Agg. Pit):
17 C = 5.8177(K/m. 25) – 25.173
R2 = 0.9839
Soil Group #3 (Sand Pit):
15 C = 3.1862(K/m. 25) + 2.5947
Data Not Used for C R2 = 0.932
13 Data Used for C Soil Group #4 (Clay Pit):
C = 33.626(K/m.25) – 69.423
R2 = 0.9556
11

9
Measured vs, Predicted Density
Measured(FHWA
vs. Estimated Density
7 Data)
160

5 150 Data
Linear Fit
140
3 130

120
1
110
-15 -10 - 5 0 5 10 15 100
y = 0.9686x
2 + 4.5858
%, Percent 90 R = 0.9383
80

ρD (GeoGauge),
ρ (HSG), pcf pcf
D

34