Sci.Int.

(Lahore),27(6),6207-6210 ,2015 ISSN 1013-5316; CODEN: SINTE 8 6207

ESTIMATION OF CALIFORNIA BEARING RATIO (CBR) FROM INDEX

PROPERTIES AND COMPACTION CHARACTERISTICS OF COARSE
GRAINED SOIL
*Attique ul Rehman1, Khalid Farooq1, Hassan Mujtaba1 and Omer Altaf2
1
Civil Engineering Department, UET Lahore, Lahore, Pakistan, 2The University of Lahore, Lahore, Pakistan
*Corresponding author [email protected]
ABSTRACT; California bearing ratio (CBR) is a useful method to assess the strength of different pavement layers by
comparing them with the strength of standard crushed rock and it is represented in percentage. Usually CBR value is used to
determine the thickness of pavement layers and also to evaluate the stiffness modulus and shear strength of subgrade material.
CBR test is quite time consuming, expensive and tedious tes,t but it is necessary to perform multiple tests throughout the length
of pavement in order to get proper idea about subgrade strength. In this regard many researchers have developed prediction
models to correlate CBR value with index properties of soil to save budget and time. This research is an effort to correlate
CBR value with classification test parameters/index properties of granular soil taken from different regions of Pakistan.
Prediction of CBR value is based upon relatively simple and less costly tests like sieve analysis and modified proctor test.
Parameters like Particle sizes at 30%, 50% and 60% passing (D30, D50, D60), coefficient of uniformity (Cu), optimum moisture
content (OMC) and modified proctor maximum dry density (MDD) are used in different combinations to develop seven
prediction models using multiple linear regression analysis. Out of seven prediction models, model-2 showed highest
regression coefficient (R2 = 0.88) and validation of model 1, 2 and 3 represented percentage error +11.9, +14.1 and +16.1
respectively. This research is only limited to granular soils (SP, SW), so the correlations developed in this research can give
better estimates only for granular soils.
Keywords: California bearing ratio, CBR, Regression, Model, prediction

1. INTRODUCTION Talukdar, 2014; Singh, Reddy and Yadu, 2011; McGough,

A strong foundation is always required for the construction of 2010 etc) [3, 4, 5, 6, 7, 8, 9,] developed prediction models
all kinds of engineering projects, especially those involving based upon index properties of soil. Present research is an
large quantities of earth works like pavements, runways and effort to develop prediction models based upon grain size
pavement embankments etc. Bearing capacity, swell pressure distribution (D60, D50 and Cu) and compaction characteristics
and settlement of different layers of pavements should be (OMC and MDD) for coarse grained soil.
within tolerable limits. Therefore, it is necessary to have
reliable methods to access the engineering properties of such 2. MATERIALS AND METHODS
projects. California bearing ratio (CBR) is one of the methods The major source of soil samples were from local deposits of
to assess the strength of different pavement layers by cohesion less soils of varying gradation namely Ravi, Chanab
comparing them with the strength of standard crushed rock and Lawerencepur Sand. A total of sixty soil samples of
and it is represented in percentage. Usually CBR value is varying gradation were prepared by mixing above mentioned
used to determine the thickness of pavement layers and also soils in different proportions, most of the samples lie in SP
to evaluate the stiffness modulus and shear strength of and SW category as per USCS classification. Three major
subgrade material. Different soil types give different CBR tests were performed on each sample, grain size distribution,
values and this test can be performed in lab as well as in the modified proctor and soaked California Bearing Ratio (CBR)
field. test. One point and three point CBR tests were performed on
CBR test is quite time consuming, expensive and tedious test few samples in order to determine a correction factor to
but it is necessary to perform multiple tests throughout the convert the one point CBR values into three point CBR
length of pavement in order to get a proper idea about values for simplicity, than only one point CBR tests are
strength of subgrade material. Many researchers developed performed on rest of the samples. The results summary
prediction models to correlate CBR value with index shown in table 2.1 is the outcome of above mentioned
properties of soil to save budget and time. As mentioned laboratory tests.
earlier, CBR value is mostly dependent on index properties of Outcomes of the laboratory tests are analyzed to develop
soil so, many investigators have conducted studies to regression models for the predict California bearing ratio.
understand the effect of soil type and soil characteristics on Multiple linear regression analysis is performed by
CBR value and also correlated CBR value with the index considering CBR value as dependent variable and D50, D60,
properties/classification test parameters. Cu, MDD and OMC as independent variables using
Breytenbach (2009) did research work to develop prediction Statistical package for the social sciences (SPSS) software.
models for estimation of CBR value using natural road Total seven different correlations are developed using
construction materials in South Africa [1]. Ferede (2010) independent variables in different combinations. Then a
developed correlations to predict CBR value using D60, OMC comparison is made between predicted and experimental
and MDD for granular soils and LL, PL, PI and F200 for fine results of CBR value using same data. In order to check the
grained soils [2]. And many other investigators like validity of developed prediction models, experimental data is
(Taskiran, 2010; Venkatasubramanian and Dhinakaran, 2011; collected from Geotechnical Engineering Laboratory of Civil
Patel and Desai, 2010; Yildrim and Gunaydin, 2011;

Nov.-Dec
6208 ISSN 1013-5316; CODEN: SINTE 8 Sci.Int.(Lahore),27(6),6207-6210 ,2015
Engineering Department, University of Engineering and 40
Technology Lahore.

Table 2.1: Summary of test results 30

GSD Modified Proctor CBR

CBR (%)
Range CBR
D30 D50 D60 OMC MDD 20
of value
results Mm mm Mm % kN/m3 %
Max. 1.1 2.3 2.8 16.6 21.92 35 10
Min. 0.17 0.22 0.25 8.1 17.64 8
0
3. RESULTS AND DISCUSSIONS 0 0.5 1 1.5
Results summary of the experimental study is shown in table
D30 (mm)
2.1, outcomes of the laboratory tests are initially analyzed to
develop regression models for the prediction of California
bearing ratio using simple linear regression analysis on Excel
Figure 3: D30 verses Soaked CBR value
as shown below in figures 1 to 6. Linear relationship among
CBR value and various parameters used are displayed below.
40

40
30

30
CBR (%) 20
CBR (%)

20
10

10
0
0 2 4 6 8 10
0 Cu
0 1 2 3
D50 (mm)
Figure 4: Cu verses Soaked CBR value

Figure 1: D50 verses Soaked CBR value 40

40
30
CBR (%)

30
CBR (%)

20 20

10
10
0
0 1 2 3 0
D60 (mm) 17 19 21 23
MDD (kN/m3)

Figure 5: Maximum dry density verses Soaked CBR value

Figure 2: D60 verses Soaked CBR value

Nov.-Dec
Sci.Int.(Lahore),27(6),6207-6210 ,2015 ISSN 1013-5316; CODEN: SINTE 8 6209
40 40
35 %age Error = + 9.9
30
30
CBR (%)

25

CBR (%)
20
20
10 15
10
Predicted
0 5 Experimental
5 8 11 14 17 20 0
OMC (%) 0 5 10 15 20 25 30 35 40 45 50 55 60 65
Figure 6: Optimum moisture content verses Soaked CBR value No. of samples

Then multiple linear regression analysis is performed by Figure 8: Comparison between predicted and experimental
considering CBR value as dependent variable and D50, D60, results for Model-2
Cu, MDD and OMC as independent variables using
Statistical package for the social sciences (SPSS) software. 40
Total seven different correlations are developed using
35 %age Error = + 11.5
independent variables in different combinations, developed
models are given below in table 2.2. 30
Table 2.2: Developed prediction models 25

CBR (%)
Eq. Developed Prediction models R2 20
15
1 0.84
10
Predicted
2 0.88 5 Experimental
0
3 0.83 0 5 10 15 20 25 30 35 40 45 50 55 60 65
No. of samples
4 0.78
Figure 9: Comparison between predicted and experimental
5 0.86 results for Model-3

6 0.87
18.0
7 0.86 15.0
12.0
CBR (%)

A comparison is made between experimental and predicted 9.0

CBR values only for model 1, 2 and 3 to get an idea about %age Error = + 11.9
percentage error/deviation in predicted CBR values. This 6.0
comparison is presented below in figure 7, 8 and 9. Predicted
3.0
Experimental
0.0
40 0 1 2 3 4 5 6 7 8 9 10
35 %age Error = + 12.8 No. of Samples
30
25
CBR (%)

Figure 10: Comparison between predicted and experimental

20
results using validation data for Model-1
15
10
5 Predicted
0
0 5 10 15 20 25 30 35 40 45 50 55 60 65
No. of samples
Figure 7: Comparison between predicted and experimental
results for Model-1

Nov.-Dec
6210 ISSN 1013-5316; CODEN: SINTE 8 Sci.Int.(Lahore),27(6),6207-6210 ,2015
2. Prediction Models developed using simple linear
18.0 regression analysis showed good R2 value but these
15.0 relationships are relatively less reliable in comparison
with multiple linear regression models.
12.0 3. Prediction Model-2 developed using multiple linear
CBR (%)

regression analysis represented relatively good

9.0
performance by showing highest R2 value of 0.88.
6.0 %age Error = + 14.1 4. Validation of model 1, 2 and 3 represented percentage
error +11.9, +14.1 and +16.1 respectively and these
3.0 Predicted errors generally fall within relatively reasonable range.
Experimental 5. This research is only limited to granular soils (SP, SW)
0.0
so; the correlations developed in this research can give
0 1 2 3 4 5 6 7 8 9 10 11 12
better estimates only for granular soils.
No. of Samples
5. REFRENCES
Figure 11: Comparison between predicted and experimental [1] Breytenbach, I.J., “The relationship between index
results using validation data for Model-2 testing and California Bearing Ratio values for natural
road construction materials in South Africa”, M.Sc.
18.0 dissertation, Department of Geology, Faculty of Natural
and Agricultural Sciences, University of Pretoria, South
15.0 Africa (2009).
[2] Ferede, Z.W., “Prediction of California Bearing Ratio
12.0
(CBR) value from Index Properties of Soil”, Department
CBR (%)

9.0 Of Civil Engineering, Addis Abab University (2010).

[3] Taskiran, T., “Prediction of California bearing ratio
6.0 %age Error = + 16.1
(CBR) of fine grained soils by AI methods”, Journal of
3.0 Predicted Advances in Engineering Software Volume 41 Issue 6,
Experimental June 2010, Pages 886-892(2010).
0.0 [4] Venkatasubramanian. C, Dhinakaran. G., “ANN model
0 1 2 3 4 5 6 7 8 9 10 for predicting CBR from index properties of soils”,
No. of Samples International Journal of Civil and Structural
Engineering, Integrated Publishing Association (IPA),
Figure 12: Comparison between predicted and experimental
Vol. 2 (2): 605-611(2011).
results using validation data for Model-3
[5] Patel, R. S. and Desai, M.D., “CBR Predicted by Index
Validity of developed prediction models is checked by using Properties of Soil for Alluvial Soils of South Gujarat”,
validation data of soil properties from Geotechnical Indian Geotechnical Conference, Proc. IGC, Vol. I, 79-
Engineering Laboratory of Civil Engineering Department, 82(2010).
University of Engineering and Technology Lahore and this [6] Yıldırım, B. and Gunaydın, O., “Estimation of California
database was not used in development of prediction models. bearing ratio by using soft computing systems,” Expert
The results of experimental and predicted soaked CBR values Systems with Applications, Vol.38, pp: 6381-
are presented in Figure 10, 11 and 12 for prediction model 1, 6391(2011).
2 and 3. This comparison is representing a close relationship [7] Talukdar, D. K., “A Study of Correlation Between
between experimental and predicted results and their California Bearing Ratio (CBR) Value With Other
corresponding percentage error is shown within presented Properties of Soil”, International Journal of Emerging
figures. Therefore CBR value can only be predicted with the Technology and Advanced Engineering, ISSN 2250-
help of developed prediction models using good engineering 2459, ISO 9001:2008 Certified Journal, Volume 4, Issue
judgment and experience. 1(2014).
[8] Singh, D., Reddy, K.S. and Yadu, L., “Moisture and
4. CONCLUSIONS Compaction Based Statistical Model for Estimating CBR
Following conclusions are drawn from presented research of Fine Grained Sub grade Soils”, International Journal
work of Earth Sciences and Engineering, ISSN 0974-5904,
1. It is concluded in this research that there is good
Volume 04, No 06 SPL, October 2011, pp 100-
relationship between index properties/classification test
parameters and soaked CBR values for coarse grained 103(2011).
soil. Therefore it is observed that CBR value increases [9] McGough, P.G., “A method for the prediction of soaked
with increase in grain size and modified proctor CBR of remoulded samples from standard classification
maximum dry density and decreases with increase in tests” Australian Geomechanics Vol. 45, No. 3, pages
optimum moisture content. 75-85(2010).

Nov.-Dec