INTRODUCTION
Density can be characterized as one of the most important parameters in the design and implementation of asphalt pavement. Selection of an appropriate compaction level is highly critical to achieve the desired performance of pavement.
After spreading asphalt mixtures on the road bed, it should be compacted by
rollers to achieve appropriate level of density. Before coming the nonnuclear
nondestructive devices, there were only two methods to determine the density
of fresh asphalt. The first, core sample method, a destructive method, is on
the basis of coring and the second method employs a portable nuclear nondestructive
device. Although, the first method is the most accurate method for determination
of density, it is costly and timeconsuming and is preferred to be applied as
the last option. In this method, samples, first, should be sent to laboratory
and since, it is likely that there is a notable distance between site and laboratory,
it may take from 1 h to 1 day. In addition to the above, when the asphalt is
still warm, the method does not provide the possibility to control the density
during different stages of rolling process (Shane et
al., 2004).
In recent years, the application of nondestructive tests has been increased,
particularly to evaluate pavement quality and control the compaction process
of pavement layers. Nuclear density gauge can be used to determine insitu specific
gravity and moisture content in accordance with ASTM D2922D3017. For this purpose,
this device employs emission of gammaray and neutron particles to compute specific
gravity and moisture content, respectively. It offers two advantages over the
constructive methods: the high speed and not being destructive. A significant
amount of studies have been conducted on the results obtained from nuclear density
gauge by different manufacturers. Site location, materials type and device model
may have appreciable effects on the results (Khani, 2004).
Nuclear density gauge is much faster and extremely costeffective compared to
core sample method, however, it is subjected to the following disadvantages:
(1) use of a radioactive source requiring high adjustment and special training
of experienced operators, (2) getting permit, renovation of equipment, technicians
training and complicated operation. In order to overcome many of the disadvantages
of nuclear density gauge, nonnuclear nondestructive device has been developed
as a safer, lighter and more costeffective alternative (Karlsson,
2002). Also Karlsson (2002) found that water content
(more than 15%) in asphalt layer effects on the instrument’s reliability.
Pedro Romero found in his study that, to measure density in the field, a calibration
procedure that is mixture specific should be carried out and also indicated
that it is necessary to correct for changes in moisture and temperature (Pedro
Romero, 2002). Robert Schmitt found in his study that nonnuclear devices
could be used for quality control, however they were not recommended for quality
assurance or acceptance testing (Schmitt, 2006). This
study is aimed at evaluating the performance of nondestructive devices using
conducting field tests and statistical analysis of obtained results.
MATERIALS AND METHODS
In order to perform field studies and evaluate the performance of two nondestructive devices, namely PQI and nuclear, some tests were performed in 2008 at the specified segments of newly established TehranPardis Freeway.
Calibration of PQI device before taking measurement operation: The role of calibration procedure in obtaining correct readings is highly critical. It should be noted that the calibration of the device for each project is necessary; therefore, Manufacture Company of PQI has presented different calibration method. In this study, normal method has been employed to calibrate the device. This is mainly due to the fact that other methods require density of unrolled asphalt mixture and prediction of the density of implemented mixture by experts as input data.
Five core samples as reference data were taken from asphalt mat and the densities of core samples were calculated. The results are shown in Table 1. The readings from PQI are represented in Table 2. From the results in Table 1 and 2 calibration process was done.
The results obtained from reading of density by PQI and coring devices are
as follows:
• 
Mean of coring results: 2181 kg m^{3} 
• 
Mean of PQI reading: 2010 kg m^{3} 
• 
Difference between coring results and PQI reading: 171 kg m^{3} 
This modification value means that 171 should be added to Mean of PQI reading
(Fig. 1).
Calibration of nuclear device before taking measurement (in asphalt layer):
The nuclear device utilized in this study is HS5001EZ and produced by TROXLER
Ltd.
Table 1: 
The obtained data from coring at five stations for device
calibration 

Table 2: 
Readings of PQI device at five stations and different local
positions (kg m^{3}) 


Fig. 1: 
Reading location layout and PQI measurement pattern 
If the difference between the results of successive measurement is minor, the device is calibrated and ready to be used.
Field measurements on the asphalt layer: In this study, 10 sections
with the width of 3.65 m that have been placed 7 m apart from each other were
selected. Afterwards, 6 points were specified at each section and finally the
density was computed for 60 points using two methods, namely nuclear and nonnuclear
nondestructive methods (Fig. 2). The reading from each method
is shown in Table 3 that in each section, 6 readings were
done by both nuclear and PQI devices and one reading was done by coring device.
Finally the density of asphalt mat was computed for 60 points to be enough for
proper statistical analysis.
Field measurements on the soil surface: The density of subgrade layer was determined at the segment of TehranPardis Freeway and the results are presented in Table 4, then, the density values obtained from sand bottle test have been utilized as a criterion for validation of outputs of nuclear device (Fig. 3).

Fig. 2: 
Test section layer for asphalt layer density measuring 
Table 3: 
Results of field measurements by three methods 


Fig. 3: 
Taking measurement for 20 points in soil sub grade layer
using nuclear device 
Table 4: 
Determination of density of soil layer using sand bottle
test 

Table 5: 
Measures of density for 20 points in soil subgrade layer
using nuclear device 

Afterwards, determination of density by nuclear device has been made at 20
points with the interval of 50 cm and the results are shown Table
5.
RESULTS
Analysis of obtained results from different methods and presenting some
practical approaches for the optimal use and reduction of error: The obtained
results from measurement of PQI and nuclear devices at the edges of segments
were lower than that of at the medium of segments. The same result was observed
in the conducted research by North Carolina Department (Shane
et al., 2004). SPSS Software has been applied to analyze data for
the purpose of validation of nondestructive devices.
The function used for conducting statistical test is called test statistic and varies in different conditions. In what follows, the relevant test statistic to each test is introduced. In parametric tests like Tstudent test, some assumptions are made concerning population distribution. For instance, Tstudent test is on the basis of this assumption that the societies will follow normal distribution. Tstudent test aimed to obtain a better understanding of unknown parameters of statistical society is normally employed to evaluate the differences between means of two societies which are under analysing. Since, the distribution of considered population is not known in nonparametric tests, accordingly, tstudent test can not be applied, because it is based on the normal distribution. In such conditions, nonparametric MannWhitney test may be used instead. The other difference between parametric and nonparametric test is that nonparametric tests employ median in lieu of mean so as to compare societies. The Ttest used to compare means of normal societies is strongly dependent on the hypothesis of equality of variances and normal distribution of statistical society. Therefore, different solutions to this problem have been proposed in the literature. One way to solve this problem is to use equivalent nonparametric tests. It should be mentioned that nonparametric tests investigate the differences between two societies through evaluation of differences between median of two societies.
Analysis for validation of PQI and nuclear devices in the asphalt layer: Comparison of two statistical societies in the parametric tests is based on the means of these two societies. Since, the considered societies in the current research follow normal distribution, thus, parametric tests (Ttest in this research) was employed to compare two societies.
If the amount of significant level (sig) in the test output is less than 0.05, it can be concluded that the assumption of equality of two societies is not valid with the probability of 95%. On the other hand, if amount of significant level (sig) is more than 0.05, then with the probability of 95%, the two societies are close to each other.
In accordance with the above results in Table 6, the measured values by PQI and coring devices did not represent significant difference with the probability of 95%. However, the measured values by nuclear and coring devices showed sensible difference with the probability of 99%.
Table 6: 
Output of SPSS software for validation of PQI and nuclear
devices using core sampling results as precise values 

Practical hints for reduction of error and optimal application of PQI device:
• 
It should be noted that PQI has appropriate capability to
identify points with improper compaction such as edges or places with insufficient
compaction 
• 
It is recommended to avoid taking measurement at the edges. This is due
to the fact that density at these points is lower than other points and
accordingly it causes some error in the calculations 
• 
Do not contact with device during measurement, because it causes some
errors in the reading of the device 
• 
Rainy weather and high moisture content of pavement may result in error
in the reading of PQI device 
• 
PQI device should be calibrated according to the type of asphalt before
starting operation 
• 
Although, there are different types of calibration methods including empirical
methods, it may be advisable based on the experiences of executive team
to employ normal which is on the basis of comparison with coring results
(the most precise method for determination of density). In other words,
since other methods require density of unrolled asphalt mixture and prediction
of density of implemented mixture by experts as input data and considered
sites often do not meet the above requirements, thus, normal method is proffered 
Validation of nuclear device to measure density of soil layers: The
sand bottle test has been run five times to properly determine density of subgrade
layer. Then, measurement was made by nuclear device for 20 times through using
direct distribution method. According to the results of sand bottle test (Table
4), it can be concluded that these data do not follow normal distribution
(five observations is not sufficient for judging whether the society is normal
or not), whereas, nuclear data follows normal distribution with the possibility
of 95%.
Since, the assumption of equality of variance of two societies can not be met with the probability of 95%, therefore employing parametric tstudent test is not justified. Hence, equivalent nonparametric tstudent test, namely MannWhitney was utilized. The main difference between parametric and nonparametric tests lies in that the assumption of normal distribution of data is not required in the nonparametric tests. In recent years, these methods have been dramatically widespread and this is because of type of society distribution which is not often given. It should be mentioned that nonparametric tests evaluate the differences between two societies through assessing the difference between the median of two societies.
The output of MannWhitney test which is aimed at evaluating the difference between median of two statistical societies (in the current study: sand bottle method and nuclear device) revealed that the value of sig (0.126) is greater than 0.05 and accordingly the assumption of median equality of two statistical societies is reasonable with the probability of 95%. Moreover, other qualified nonparametric tests also confirmed the above finding.
In Table 7, the validation of the obtained results from nuclear
device using nonparametric MannWhitney, Moses and Kolmogorove tests are illustrated.
In all tests, it is assumed that null hypothesis implies equality of medians
of two statistical societies (the results of sand bottle and nuclear methods)
and the opposite hypothesis represents nonequality of medians.
Table 7: 
Output results relevant to nonparametric mannwhitney, moses
and kolmogorove tests as alternatives for tstudent test 

Hence, the obtained
results indicated that nuclear device can be considered a reliable method to
determine density of soil layers.
DISCUSSION
Based on the results of statistical hypothesis testing in this study, use of
the PQI model 301 for both Quality Control (QC) and Quality Acceptance (QA)
testing is sufficient but it found that calibration should be done to achieve
accurate results. Numerous studies on the suitability of nonnuclear pavement
density gauges for QC and QA testing of pavements were reviewed. Most of these
studies have concluded that nonnuclear gauges are suitable for contractor QC
but not recommended for QA. Also It was found that the PQI appears to be an
acceptable alternative to the nuclear gauge for density profiling, even if not
calibrated (Sargand et al., 2005). A study prepared
by the University of Utah indicated that, Calibration of these devices to local
materials and conditions is critical to obtain accurate results. Whenever practical,
the calibration should be done using a test section (Pedro
Romero, 2002).
The following conclusions are drawn from the results of this study:
• 
According to the results of validation tests for PQI and Nuclear
devices, indicated that PQI device has sufficient reliability to determine
density of asphalt mixture but nuclear device is not reliable to determine
density 
• 
Due to the negative difference between mean measurements of nuclear and
coring devices, it can be concluded that the measured values by nuclear
method are lower than that of by core sample method with the probability
of 99 % and include zero. While, this difference for nuclear and core sample
method does not include zero and is negative. Hence, the validation results
of PQI and nuclear devices indicated that contrary to nuclear device, PQI
device has sufficient reliability to determine density of asphalt mixture 
• 
The obtained results from validation of nuclear device revealed that it
has sufficient reliability to determine density of soil layers 
• 
It should be noted that PQI has appropriate capability to identify points
with improper compaction such as edges. It is recommended to avoid taking
measurement at the edges. This is because the density at these points is
lower than other points and accordingly it causes some error in the calculations 
• 
PQI device should be calibrated according to the type of asphalt before
starting operation 
• 
Although, there are different types of calibration methods including empirical
methods, it may be advisable based on the experiences of executive team
to employ normal which is on the basis of comparison with coring results
(the most precise method for determination of density). In other words,
since other methods require density of unrolled asphalt mixture and prediction
of density of implemented mixture by experts as input data and considered
sites often do not meet the above requirements, thus, normal method is proffered 
• 
The required time for calibrating the Nuclear device is much affected
by type of project and used materials (asphalt, soil) 
• 
Finally, it is also noteworthy that the most precise and reliable method
to determine the pavement layers density is core sample method 