CHAPTER 5 - UNPAVED ROAD GRAVEL
LOSS ANALYSIS
Digitised by the University of Pretoria, Library Services, 2012
Digitised by the University of Pretoria, Library Services, 2012
79
5. 1
SCOPE OF THE GRAVEL LOSS STUVIES
Regravelling is the major maintenance operation on unpaved
roads,
and it is analogous in importance to overlaying a paved road
with asphaltic concrete.
It is
therefore
important that the
ag~ncy
responslble for regravelling know when it should be programmed.
gravel loss studies were aimed at predicting the
The
loss of surfacing
material on sections with laterite and quartzite gravel and sections
i.e., clay.
without gravel,
The surfacing material of these
latter
sections contained more than 35 percent material passing the 0.074 mm
sieve.
A data summary of the dependent and independent variables
studied is given in Table 5.1.
These statistics aid in putting the
model inference space into perspective.
in Working Documents 9,
Queiroz,
13,
14,
Original data are contained
and 15 of this project
(Visser and
1979).
APPROACH FOR GRAVEL LOSS ANALYSIS
5.2
Grave 1 1 o s,s i s de f i n e d as the change in grave l
over a period of time.
gravel level or gravel
t hi c k n e s s
On a well compacted subgrade the change in
height is
the change in gravel thickness.
Although gravel thickness is not necessarily equivalent to gravel
level or gravel height under all conditions,
in this report
gravel thickness is
as a synonym for gravel level or height.
loss is a change of gravel thickness over time,
used
Since gravel
it was not necessary
to determine an absolute value at some initial point in time as was
done for unpaved roughness
and rut depth.
for the interval between regravellings,
sis cycle,
Gravel loss was evaluated
which initiated a new analy-
or from the time of the first observation until a regrav-
elling occurred.
Three major influences were identified as affecting gravel
1o s s .
T h e s e a r e we a t h e r i n g ,
the form of blading.
t r a f f i c , and the in f 1u en c e of maintenance i n
Material properties ancJ road alignment and width
then influence the gravel loss generated by each of these influences.
The general model is then:
gravel loss=
(t-ime)fl + (time)(average daily traffic)f 2
(bladings)f3
+
Digitised by the University of Pretoria, Library Services, 2012
TABLE 5. 1 -
UNPAVED ROAD DATA SUMMARY
(Continued)
00
0
(1/Rad)
Road width
Maximum
on curved sections
( m)
12. 0
7. 0
1 . 09
9. 8
.0055
.0025
.0009
.0039
8.2
0. 0
2.6
3.8
(% )
Curvature
Minimum
= 48
Number of sections
Grade
Ran e
Standard
Deviation
Mean
Variable
MATERIAL PROPERTIES
Percentage passing the 0.42 mm sieve
53
22
24
98
Percentage passing the 0.074 mm sieve
36
24
10
97
Plasticity index
11
6
0
33
32
9
20
62
88
64
11
288
7
7
0
29
Pickups
37
29
4
11 5
Two axle trucks
56
93
1
4 35
Liquid limit
(%)
(%)
AVERAGE DAILY TRAFFIC
(both directions)
Passenger cars
Buses
Trucks and trailer combinations with more
than 2 axles
TIME RELATED INFORMATION FOR GRAVEL LOSS
Number of observations
· Time of observation relative to start of
observation or regravelling (days )
I
15
I
604
I
2 38
I
18
I
0
I
66
I
211
I
0
I
1099
I
0
I
23
Number of bladings relative to start of
observation or regravelling
I
2. 3
I
3.3
Digitised by the University of Pretoria, Library Services, 2012
TABLE 5.1
- UNPAVED ROAD DATA SUMMARY
Variable
(C onc lu sion )
Mean
Standard
Deviation
Range
Minimum
Maximum
INFORMATION RELATED TO ROUGHNESS MEASUREMENTS
Roughness
(QI* counts/km)
1 17
61
75
70
16080
17880
15
44 5
Number of days since blading for the last
observation in each blading period
661
Number of vehicle passes since blading for the
last observation in each blading period
63
136460
0
75
INFORMATION RELATED TO RUT DEPTH MEASUREMENTS
Rut Depth
(mm)
Number of days
11 . 1
si~ce
8. 6
blading for the last
observation in each blading period
61
66
12490
14030
661
Number of vehicle passes since blading for the
last observation in each blading period
21
86700
co
Digitised by the University of Pretoria, Library Services, 2012
82
where f
1 •
f
2 •
and f
3
are
linear
co~binations
of material properties
and road alignment and width.
The average elevation of a subsection relative to the bench
mark was
used to evaluate gravel
ed at about
three monthly intervals,
it was
shown
These elevations were obtain-
and it was not possible to sepa-
In the Kenya study
rate seasonal influences.
1 9 7 5)
loss.
(Hodges,
Jone~
Rolt and
that no seasonal pattern existed in the data.
and
Furthermore.
this also appeared to be the case for the Brazil data.
seasonal influences do not have any practical implications since the
agency responsible for regravelling wishes to know its frequency
terms of years.
in
and has little interest in the influences of each par-
ticular season.
ANALYSIS OF GRAVEL LOSS
5. 3
The analysis of gravel
pendent variables:
since regravelling;
limit and
(5)
(1)
(2)
loss considered the following inde-
time in days since observations started or
grade;
(3)
horizontal curvature;
plasticity index of surfacing material;
centage of surfacing material passing the
0.074 mm sieve;
~.g.,
(8)
buses,
t h ·e
0.42 mm sieve and
per -
(7)
the
qualitative description of the surfacing type.
laterite and quartzite;
each of cars.
(6)
( 4) liquid
pickups.
truck-trailer combinations;
(9)
numbers of vehicles per day of
two-axle trucks and other trucks and
(10)
road width;
and
(11)
the number of
bladings since observations started or since regravelling.
factor interactions were also investigated.
i.~.,
Three
time and bladings
times two factor interactions of the other independent variables.
The GLM procedure of the SAS statistical package
tute,
1979) was
used to determine the significant factors.
cedure permitted evaluation of different
(SAS InstiThis pro-
combinations of factors,
un-
like stepwise regression where it is difficult to determine significant effects in cases of high correlation among factors.
Two models containing terms multiplied by the number of
bladings were developed within experimental conditions.
org~nizations.
where the ma-
ximum number of bladings was 23.
Some
U.S.
blade a road at very frequent inter-
Forest Service (Lund,
1973),
such
as the
Digitised by the University of Pretoria, Library Services, 2012
83
vals- some times daily- and this gives a value of B far in excess of
that used to develop the model.
gravel
loss predictions.
This leads to
To overcome this
unrealistically high
limitation a model contain-
ing only interactions with time was investigated.
The significant
factors are the following:
GL
0(-1.58
+
0.366 G
+
0.0132 NC
+
+
0.083 SV- 0.210 PI
0.0081
NT
+
420.45/R)
(5.1)
where
GL
gravel
thickness
loss in
0
t i me p e r i o d c o n s i d e r e d ,
G
absolute value of grade in percent
SV
percentage of surfacing material passing the 0.074 mm
m~
i n h und r e d da ys , i . e . ,
d a y s I 1 0 0;
sieve;
PI
plasticity index
(PI)
NC
average daily car and pickup traffic,
NT
average daily truck traffic,
R
radius of horizontal curvature,
both directions;
both directions;
in m.
The t-values of each coefficient are given in Table 5.2.
The R-squared of this model is 0.60,
standard error of
residuals,
GL
+
the
model
the sample size
11.43.
the approximate 95 percent
604,
and
the
Assuming normality of the
confidence
interval
is
or - 22.8 mm.
Several observations relate to this model.
1.
Increasing the grade,
the percentage of material pass-
ing the 0.074 mm sieve,
traffic.
gravel
2.
the average daily car and truck
or decreasing the radius of curvature increases
loss.
Increasing the plasticity index decreases the gravel
1o s
s , i . e . , t he p .1 a s t i c i t y i n de x rep re se n t s
or binding ability of the fine
3.
material.
Gravel
loss
car is
twice that of one truck passage.
taken in
t h e c e me n t i n g
associated with the pass·age of one passenger
Care should be
attaching too much weight to the vehicle equiv-
Digitised by the University of Pretoria, Library Services, 2012
84
TABLE
5.2
-GRAVEL LOSS
Parameter
D
REGRESSION
ANALYSIS
(MODEL 5.1)
Standard
Deviation
Estimate*
t-value
1 . 58
0.96
1 . 64
D
X
G
0.366
0 . 103
3.56
D
X
sv
0.083
0. 016
5.01
D
X
PI
-0.210
0.054
-3.88
D
X
1\JC
0.0132
0.0030
4.40
D
X
NT
0.0081
0.0027
3. 0 1
D/R
* Negative
420.45
sign
denotes
gravel
116.07
3.62
loss.
Digitised by the University of Pretoria, Library Services, 2012
85
alency factors,
4.
as demonstrated below.
Attempts were made to evaluate the
the different vehicle types,
numbers of buses,
pickups
effects of each of
but the relatively small
and trucks other than two-
axle resulted in insignificant influences,
or these
vehicle influenc es resulted in illogical signs on the
coefficients which were incompatible with field experience.
5.
Consequently only two
vehicle types were
used.
Surfacing material properties explain e d the influence
of surfacing type sufficiently well such that the
qualitative surfacing type descriptors were not found
to be significant .
Model
(5.1)
explicitly defined.
also contains blading influences although not
When using this model it is
assumed that the
importance of blading influences decreases as the frequency of blading
increases.
loss
For example,
the effect of a blading every day on gravel
is not the same as a blading every three months.
it is assumed for Model
(5.1)
that
On the average,
blading influences are consta-nt
irrespective of the number of bladings that occur.
5.4
DISCUSSION OF THE MODELS
Despite of the dangers of developing vehicle equivalency
f act or s ,
model
Mo d e 1 ( 5 . 1 ) ha s a wi d e a p p l i c a b i 1 i t y a n d i s s u g g e s t e d a s
for general
Th e
use.
effects of the factors
be significant are in general agreement with field
com pari so n of gravel
for t wo
the informa t ion for Section 205,
centered through
25 1,
that were found
experience .
A
5 . 1 a nd 5 . 2 .
which,according to
Fi g ur e
the
5 . 1 s h o ws
records,
was
The predicted gravel
loss was
the mean date and mean gravel thickness.
Section
an
to
loss measurements and the predicted gravel loss
sections are g i v e n i n Figure s
never bladed in
t he
18 month period.
on the other hand,
was
heavily trafficked,
received frequent
bla dings and is one of the sections on which almost three years of
data were collected.
the mean gravel
grave llings.
The predicted curves were again
thickness
Both figures
s urements and predictions.
centered through
and mean time for each period between red e monstrate good concordance between meaIn some cases
the gravel thickness
mea-
Digitised by the University of Pretoria, Library Services, 2012
00
0"1
~
~
30
~
20
L
0
•
•
'
0
[2
liJ
n..
;z
0
~
/
cr
w
(/)
co
•
•~
I
0
2
GRAVEL LOSS
PER YEAR)
10
:::>
(!)
PREDICTED
( 20.4 nrn
0
0::
77.5
:;:)
0
78 . 0
I
......
•
-~.0
78 .5
DATE
Q
I..U
:I:
z
<I
-10
I
•
•
79.5
LEGEND
UJ
i -20~
I
•
-
o -
SUBSECTION WITH MAINTENANCE
SUBSECT iON WITHOUT MAINTENANCE
•
-3Ql
FIGURE 5.1- PREDICTED AND
MEASURED GRAVEL LOSS ON SECTION
••
205.
Digitised by the University of Pretoria, Library Services, 2012
x
0
I
I
01
9NITI3AV~93~
(f)
(f)
og
S?>
ILI....J
I-ILI
~(l:
0 ct
a..<!>
I
I
I
I
I
~I
9NIT13AV~3~
x
I
I
I
I
I
I
I
I
I
I
10
I
I
I
I
I
I
I
I
0
I
0
I
X
A~V~.L18~V WOH::I
9NI1l3AV~3~
l3A3l
0
~
0
)(
0
s
( uw) SS3N~~IH.L
(f)
~
z
w
8
0
~
z
(f)
:::l
(/)
(I)
i=
0
u
w
i=
u
ILl
(f)
(I)
:::>
(f)
)(
I
I
I
I
I
co
0
I
l 9b
d
m
0
l3AV~9
I
I
I
o,
I
0
1'--
0
0
a)
1'--
0
co
1'--
ILl
1<l:
0
N
....
It)
z
0
i3
L&.l
z
(I)
0
(I)
(I)
9
4(
z
Q
Q
L&.l
...u
0
1&.1
I
f
N
lri
1&.1
G:
(!)
:l
ii:
87
Digitised by the University of Pretoria, Library Services, 2012
88
sured after regravelling
er than that predicted.
atively
on,
for example, Section 251,
This is
well-compacted as shown by the
ness,
although
attributed to the material being rel-
In other cases,
uncompacted.
relatively
predicted and measured gravel thick-
but since this is extremely variable
little information on the degrees of compaction is
able,
this will have
tities,
used
It could be argued that it would be necessary to pre-
diet initial compaction effects,
to
the material was
to the best knowledge only traffic compaction was
in all cases.
and
was much hi gh-
achieve a final
When calculating gravel quan-
little benefit .
engineers take
b ulking
usually avail-
and compaction effects into account
compacted thickness .
In repeating this
study it may be worthwhile to i n i t i a t e measurements,
say,
type of
one month
after regravelling, to overcome initial compaction effects.
T a b l e 5 . 3 wa s
ing some feel
g e n e r a t e d f r o m Mo d e l
for the behavior of the model.
( 5 . 1 ) t o a s s i s t in obtain Values covering the
This
extremes of the independent variables were selected.
in combinations which normally do
not exist in practice,
resulted
such as a
surfacing material containing 10 percent of the material passing the
0 . 074 mm sieve which has a plasticity index of 30.
model predicts a gravel gain,
diction
which is
unrealistic.
uf unrealistic values for cases such as
s a r y t o us e
a n o mi
gravel loss is
il
al
less
a nn ua l
t han
gr a vel
thi s
For this
l os s of ,
t he
value .
since it resulted in
to held experience,
it is neces-
5 mm w h e n e v e r
An attempt to include cumulative rainfall
was un s uc c ess f ul
To avoid the pre-
these,
say,
case the
in
the
model
a model that predicted contrary
and al s o re s ulted in a change-over of effects
within the data range studied .
5.5
SUMMARY
Model
of the
(5.1)
is
recommended for general
predictions from this model with data collected in Kenya show
good agreement.
From a very
though the Brazil data was
annual
use and comparison
rainfall,
the
low rainfall regions,
include rainfall
in
collected in a region of about
mod e l
~.g.,
th e
limited comparison it appears that al-
is
1600 mm
valid for predicting gravel
below 750 mm per year.
Brazil gravel
loss in
An attempt to
lo ss prediction model was
un-
Digitised by the University of Pretoria, Library Services, 2012
0
·0_>,
TANGENT
9~
.
0
10
180
400
0
400
15
400
15
400
15
400
15
400
9.5*
28.1
21.3
39.9
18 .l
36.6
29.9
48.4
0
90
33.8
52.3
45.6
64.1
42.3
60.8
54.1
72.7
10
- 13.5
5.1
-1.6
16.9
- 4.9
13.6
6.9
25.4
90
10. 1
29.3
22.6
4]
.1
19.3
37. 8
31.1
49.7
10
20.2
38.8
32.0
50.6
28.7
47.3
40.6
5 9.1
90
44.4
6 3.0
56.3
74.8
53.0
71.5
64.8
8 3. 4
10
- 2.8
15.8
9.0
2 7.6
5.7
24.3
17.6
36.1
90
21.4
40.0
33.3
51.8
30.0
48.5
41.8
60.4
0
30
0
8
30
* OBS:
POSITIVE
VALUES
DENOTE
GRAVEL
LOSS
TABLE 5.3- ANNUAL GRAVEL LOSS GENERATED FROM MODEL 5.1.
00
\.0
Digitised by the University of Pretoria, Library Services, 2012
90
successful.
Under certain combinations of factors outside the range in
which the model was developed,
gravel loss can occur.
gravel
gravel
To overcome this problem a minimum annual
loss of 5 mm should be
dictions.
gravel gains or unrealistically low
used,
Very short radius curves
loss predictions,
irrespective of the model preresult in
and from a comparison with the Forest Service
Study a substitution of a 100 m radius
Model
(5.1)
is
unrealistically high
for smaller radius curves in
recommended.
Digitised by the University of Pretoria, Library Services, 2012