INTRODUCTION
Porous asphalt is described as a bituminous bound mix with selected grading
and highquality aggregates to provide an asphalt mix with 20% air voids. The
presence of continuous air voids in the wearing course allows water to pass
through the mix laterally and eliminates ponding water (Alvarez
et al., 2006). Other benefits include elimination of glare, splash
and spray, improvement in visibility, increased skid resistance of the road
surface and improved traffic safety especially in wet conditions (Moore
et al., 2001; Cooley et al., 2000;
Poulikakos et al., 2004).
Porous asphalt also offers vast benefits to the environment. The good sound
absorption potential has led to the application of porous asphalt on the expressway
close to residential areas to attenuate noise. Alvarez et
al. (2006) mentioned that by using porous asphalt, noise level can be
reduced by approximately 3 dB (A) compared to dense asphalt. Higher reduction
of 14 dB (A) was recorded on roads in Texas when asphaltrubber open grade friction
course was used to overlay an existing continuous reinforced concrete pavement
(CRCP). Pagotto et al. (2000) pointed out that
porous asphalt had also improved the quality of water runoff where heavy metal
loads discharged into the environment were reduced with approximately 20% cuprum
(Cu), up to 74% plumbum (Pb); solids were detained at a rate of 87% and hydrocarbons
were intercepted at an even higher rate of 90%. It was reported that higher
percentage of pollutants were found filtrated through porous asphalt from storm
water which contained 99% of suspended solids, 38% phosphorus, 96% zinc and
99% of petroleum hydrocarbons from diesels (Roseen et
al., 2007).
In Malaysia, porous asphalt was laid primarily to improve traffic safety. In
general, Malaysia experience wet and humid tropical climate and subjected to
high rainfall intensity throughout the year. Poor drainage system can result
in hydroplaning especially for fast moving traffic. Under such circumstances,
contact between tire and pavement can be eliminated. This hazardous situation
leads to loss of control for braking and steering while driving. According to
Hamzah et al. (2004), porous asphalt was first
initiated on Malaysian road along the CherasBeranang Road in 1991. Subsequently,
numerous patchy applications of the material were currently laid along the NorthSouth
Expressway and elsewhere. The use of porous asphalt pavement in Malaysia had
resulted in a reduction of the number of accidents and promotes safety to the
road users especially during wet weather. However, its service life is somewhat
limited due to poor durability compared to conventional dense mix. Clogging
is also a major problem that occurred throughout the service life of porous
asphalt due to the open nature of the mix itself. Various types of clogging
agent that may adversely affect the permeability of mix include tyre wear byproducts,
dirt, road dust, residual soils deposited from dirty wheels and lorries transporting
earth. According to Hamzah et al. (2010), the
abrasive action of vehicle wheel on pavement surfacing, especially on high stressed
areas, can initiate particle loss while the action of water results in stripping.
This leads to a pavement distress type known as ravelling and is more predominant
in porous asphalt compared to conventional mix.
The national specifications for porous asphalt first appeared in Malaysia in the year 2008 when the Public Works Department (PWD) launched the specifications on specialty mixes that includes porous asphalt. Two porous asphalt gradations, designated as Grading A and Grading B in this study, are specified and they differ in terms of their nominal maximum aggregate sizes, which is 10 mm and 14 mm, respectively. The Malaysian PWD also suggests the use of modified bitumen and polymer additive in 70/100 pen grade bitumen to improve durability. Based on the new specifications, this study was carried out to evaluate the performance of both porous asphalt in the laboratory. In addition, a limiting value on the resistance to disintegration was set to suit typical ambient temperature in Malaysia using a Curve Estimation Regression based on established literature reviews. For precise experimental data interpretation, the General Linear Model (GLM) Univariate through Analysis of Variance (ANOVA) was used to identify whether there was any substantial effects of aggregate grading, bitumen type and bitumen content on the properties of porous asphalt mix. Subsequently, two types of bitumen were incorporated in the preparation of both mixes. Mix performance was quantified in terms of permeability, air voids, abrasion loss, and Indirect Tensile Strength (ITS). From statistical analysis, gradation, binder type, binder contents and the interactions between the factors has a significant effect on the properties of porous asphalt mixes tested.
MATERIALS AND METHODS
Materials: The experimental work involved used of crushed granite aggregates produced by Kuad Quarry Sdn. Bhd in Penang on December 2008. Crushed granites were washed, dried, and sieved into the selected range of sizes according to the Malaysian porous asphalt gradations as shown in Fig. 1. Table 1 shows the basic properties of crushed granite aggregate. The result shows that all aggregate properties satisfy the PWD specifications. Conventional bitumen 60/70 pen grade and PG76 modified bitumen were used for the whole specimen preparations as well as hydrated lime and ordinary Portland cement as the fillers. The properties of binders are presented in Table 2. Characterization of the materials properties were conducted in the Highway Engineering Laboratory, Universiti Sains Malaysia.
Mixing and compaction temperatures: The mixing and compaction temperatures
are dependent on binder viscosity as determined from a Brookfield Viscometer.
According to the Asphalt Institute (2007), the ideal mixing
and compaction temperatures correspond to conventional binder viscosities of
0.17±0.02 and 0.28±0.03 Pa.s, respectively. However, Yildrim
et al. (2006) reported that from his study, the ideal mixing and
compaction temperatures for modified binder are optimal at viscosity of 0.275±0.03
Pa.s and 0.550±0.06 Pa.s. Figure 2a and b
display the viscositytemperature relationships for both binders. The solid
line indicates the graphical plot between viscosity versus temperature, while
the temperature ranges adopted for mixing and compaction are represented by
the dashed and dotted line respectively and abbreviated in Table
3.
Table 1: 
Properties of crushed granite aggregate used in this study 


Fig. 1: 
Aggregate gradations 

Fig. 2: 
Temperature and viscosity relationship. (a) 60/70 and (b)
PG76 
Table 2: 
Properties of asphalt binder 

Table 3: 
Mixing and compaction temperatures adopted 

Specimen preparation: Each specimen comprised of about 1100 g of aggregate
batches in addition to asphalt binder. The aggregates and asphalt binder were
blended at their corresponding mixing temperatures. The loose hot mixes were
conditioned at the compaction temperature for 2 h as recommended by the Asphalt
Institute (2007). Compaction followed subsequently by applying 50 blows
on each face of the specimen using the standard Marshall hammer. The compacted
specimens were then subjected to a series of performance tests. Specimens were
designated based on their gradations as shown in Table 4 for
ease of reference.
Air voids determination: Prior to the air voids determination, the bulk
specific gravity of a compacted mix (G_{mb}) was ascertained using the
Specimen Geometry Method (SGM) by measuring its diameter (d), thickness (H)
and the mass of the specimen in air (M_{a}). The theoretical maximum
density (G_{mm}) of the loose mix was determined according to ASTM D2041
procedure (ASTM, 2000). Air voids (V_{a}) in
the compacted specimen was calculated by using equations derived from the Asphalt
Institute (2007).
Table 4: 
Specimen designation 

To ensure the porous asphalt has sufficient permeability, the air voids contents
should be greater than 18% (JKR, 2008).
Water permeability test: Hydraulic conductivity is an important property
of porous asphalt. The hydraulic conductivity of the compacted specimens was
expressed in terms of the coefficient of permeability (k), and determined using
a falling head water permeameter. A typical test on asphalt sample cured at
ambient temperature involved pouring water into the perspex tube and sealing
the orifice with the rubber bung. Once steady state conditions have been reached,
the rubber stopper was then removed and the time taken for water to fall between
two designated points on the perspex tube was noted. According to Mallick
et al. (2000), permeability greater than 0.116 cm sec^{1}
(100 m day^{1}) is recommended to ensure a permeable mix.
Cantabrian test: The Cantabrian test was conducted to determine the
abrasion loss property of the mix using the Los Angeles drum without steel balls.
A specimen was placed in the drum and rotated for 300 revolutions at 3033 rpm.
The specimen was subjected to abrasion and impact forces during the rotation
of the drum. The masses of the specimens before and after the test were recorded.
According to Herrington et al. (2005), the Cantabrian
test is usually carried out at 25°C in many countries worldwide.
Table 5: 
Overview on limiting value of abrasion loss at various temperatures 

This may be particularly true for temperate countries during the hot season.
However, for countries in the tropics, the ambient temperature is higher than
25°C and this includes Malaysia. Maintaining the specimen temperature at
25°C throughout the Cantabrian test under local conditions would be an arduous
task. Therefore, some adjustments were made and the abrasion loss criteria were
carried out to determine the limiting abrasion loss value at a typical Malaysian
ambient temperature of about 30°C. Prior to the test, the specimen was conditioned
at 30°C for 4 h. To predict the limiting abrasion loss at 30°C, an extrapolation
of permitted abrasion loss from the values given in Table 5
was necessary.
Determination of curve model: The best fit model of abrasion loss was established using the Curve Estimation Regression in SPSS 11.5 software. Four basic models have been chosen for comparison, namely linear, logarithmic, power and exponential.
Indirect tensile strength test: The Indirect Tensile Strength (ITS)
was used to determine the splitting strength and serves as an indicator of the
tensile strength of the mix. In the test, the specimens were subjected to compression
load which acted parallel to the vertical diameter plane by using the Marshall
testing machine. Kok and Yilmaz (2009) described the
failure of the specimen by cracking along a loaded plane where tensile stress
loading acts perpendicular to the specimen. Preceding the test, specimens were
cured at 25°C in an incubator for 4 h. The test was done in accordance with
ASTM D4123 procedure (ASTM, 2005).
RESULTS
Best fit curve model: Table 6 indicates the result of Curve Estimation Regression on linear, logarithmic, power and exponential models. From the graphs plotted, the power model gives the highest coefficient of determination (R squared) with an accuracy of 0.979. Therefore, it was selected as the best curve to determine the limiting abrasion loss value at 30°C.
Figure 3 shows the selected power model relationship of permitted
abrasion loss at various temperatures obtained from Colonna
(1996). The permitting abrasion loss value at 30°C extrapolated from
this curve equals to 16%.

Fig. 3: 
Abrasion loss limiting value 

Fig. 4: 
Air voids results 
Therefore, this value was adopted as the reference point to determine the
minimum permissible abrasion loss. The solid line indicates the relationship
between abrasion loss and test temperatures. The dashed line represents the
reference point for the limiting abrasion loss at 30°C.
Air voids: The air voids of both mixes and four different binder contents are estimated based on the bulk specific gravity (G_{mb}) and maximum theoretical density (G_{mm}). Figure 4 shows the typical trend of decrement in air voids as the binder content increases. The air voids of Mix B is higher than Mix A. The air voids for Mix B vary between 20.7 to 22.8% compared to 16.4 to 20.7 % for Mix A.
This is due to the fact that Mix B contains higher Nominal Maximum Aggregate
Size (NMAS) which is 14 mm compared to 10 mm in Mix A. Mix A contained almost
20% higher aggregate size passing the 5 mm sieve size compared to Mix B.
Table 6: 
Curve estimation of abrasion loss values from literature
review 


Fig. 5: 
Permeability test results 
Air voids is also influenced by the fine aggregate content in the gradation.
The minimum recommended air voids is 18% as shown by the horizontal dashed line.
Based on this criterion, only Mix B satisfies the criteria for all bitumen contents.
Coefficient of permeability: The effects of permeability on porous asphalt were evaluated for both mixes. The test results are shown in Fig. 5 and the dashed line indicates the recommended coefficient of permeability to ensure good drainage capacity. The result implies that there is a reduction in the coefficient of permeability for both mixes as the binder content increases. Similar pattern of air voids are found for both mixes, whereby Mix B exhibits higher coefficient of permeability compared to Mix A because the former has higher air voids. However, all mixes satisfy the recommended minimum value except for the A6 and AP mixes prepared at 5% binder content.
Abrasion loss: Figure 6 illustrates the results of the Cantabrian test. The recommended permitted abrasion loss at 30°C is indicated by the dashed line. All mixes are found to satisfy the limiting value stipulated in the specifications except for Mix A6, AP and B6 at 3.5% binder content. By comparing the results of both mixes, the results reveal that the abrasion loss of Mix A is higher than Mix B. The use of modified binder increases mix resistance to abrasion loss for both mixes, the corresponding values are 35.2 and 51%, respectively. In addition, the abrasion loss of both mixes reduces as the binder content increases.
Indirect tensile strength: Figure 7 shows the mean
ITS of the mixes at different gradings, binder types and bitumen contents.

Fig. 6: 
Abrasion loss results 
When gradation is taken as the variable, it is clearly seen that the ITS of
Mix A is 85.6% higher compared to Mix B. This implies that the higher air voids
in Mix B result in reduced contact area between aggregate particles, hence leading
to its lowered ITS. However, the use of modified bitumen increases the mean
ITS values of both mixes by 18.1%.
DISCUSSION
General Linear Model (GLM) Univariate analysis was used to identify whether there was any substantial effects of aggregate grading, bitumen types and contents on the properties of porous asphalt mix. GLM Type III Sum of Squares for the range of binder contents from 3.5 to 5% with a confidence level of 95% (α = 0.05) was adopted.
Air voids: Table 7 presents the results of GLM analysis
which shows the effects of Agregate Grading (GT), Bitumen Type (BT), Binder
Contents (BC), and interactions between the factors on the air voids. However,
not all factors have significant effects on the air voids.
Table 7: 
Single and interaction effects on the air voids 

^{a}R Squared = 0.979; Adjusted R Squared = 0.959 
Table 8: 
Single and interaction effects on the coefficient of permeability 

^{a}R Squared = 0.953; Adjusted R Squared = 0.909 
The factors that seems to have no effects at all on the air voids of porous
asphalt are single factor (BT) and interaction factors (BT*BC and GT*BT*BC)
which exhibit a pvalue greater than 0.05. This implies that the application
of different binder types is not significant to improve the air voids of porous
mix. The results of this study showed that incorporationg different binder content
influence the air voids of porous asphalt mixes. This is similar to the findings
of Faghri and Sadd (2002) but there was no statement
of statistical significance. Mogawer et al. (2002)
found that gradation have a significant effect on the air voids of asphalt mixes.
Coefficient of permeability: Table 8 shows the analysis
of results corresponding to GT, BT and BC effects on the coefficient of permeability.
The results indicate that all the single factors have significant effects but
the combination factor shows otherwise. However, there is a strong correlation
on the analysis due to higher R squared value of 0.953. It implies that the
model fits the data very well. From a study done by Mogawer
et al. (2002), it was indicated that gradations have significant
effects on the permeability of hot mix asphalt mixes. Another study by Faghri
and Sadd (2002), reported that the use of different types of binder had
increased the permeability of porous asphalt mixes.
Abrasion loss: The GLM analysis was performed on the abrasion loss results
as shown in Table 9. The results indicate that the gradation
(GT) does not contribute to any substantial effects on the reduction of abrasion
loss value for each binder contents. Furthermore, all interaction factors do
not give any significant effects on the changes in abrasion loss of porous asphalt
mixes tested.
Table 9: 
Single and interaction effects on the abrasion loss 

^{a}R Squared = 0.971; Adjusted R Squared = 0.943 
Table 10: 
Single and interaction effects on the indirect tensile strength 

^{a}R Squared = 0.994; Adjusted R Squared = 0.988 
However, Nielsen et al. (2004) and Watson
et al. (2003) reported that the use of modified binder had improved
the resistance of porous asphalt to particle loss. Suresha
et al. (2009) states, that binder content has significant effect
on the abrasion loss of porous friction courses.
Indirect tensile strength: Table 10 presents the
results for the GLM on the ITS results. The results indicate that all the factors
are significant in affecting the ITS values, except for the interaction factor
of GTxBTxBC which has a pvalue (0.81) higher than 0.05. A study done by other
researchers had also found that the use of polymer modified bitumen has increased
by nearly doubled the strength of opengraded mixes (Faghri
and Sadd, 2002).
CONCLUSIONS
This study presents the result of the new Malaysian porous asphalt mixes prepared according to gradations A and B. Generally, Mix B exhibits better performance in terms of permeability, and resistance to abrasion loss compared to Mix A, but the ITS results shows otherwise. The use of modified binder significantly increases the resistance to abrasion loss and ITS of the mixes. The GLM analysis provides a better understanding on which factors that have a significant influence on the performance of porous asphalt. This model can be used as a guide to improve the properties of porous asphalt with consideration to the related factors.
ACKNOWLEDGMENTS
The authors would like to acknowledge the Malaysian Ministry of Science, Technology and Innovation that has funded this research grant through the Science Fund program that enables this paper to be written. Many thanks are also due to technicians of the Highway Engineering Laboratory at the Universiti Sains Malaysia for their help. Acknowledgments are also due to Shell Ltd. and Kuad Quarry Sdn. Bhd., Penang respectively for the supply of bitumen and aggregate