Performance of a Sensitivity Analysis on the Multi-Function Network Survey Vehicle (MFNSV)

Multi-Function Network Survey Vehicle (MFNSV) provides cutting-edge technology for obtaining highway pavement information such as; roughness, rutting, geometric data, road condition data as images and accurate distance measurements. The Hawkeye processing toolkit is used to process the data collected through a moving vehicle. The MFNSV manufacturer claims that the information is fairly accurate when the vehicle is driven within a speed range of 20 to 110 km/h. Road pavement information was collected manually through accurate field measurements along 4 sample stretches on the A004 road, and the same information was collected through the MFNSV. The results obtained from the two methods were compared by carrying-out a sensitivity analysis test. The results of the sensitivity analysis showed that the information collected from all 4 sample test sites did not possess significant statistical differences between the data collected by the two methods.


Introduction
It was intended to conduct a sensitivity check on the Multi-Function Network Survey Vehicle (MFNSV) (see Figure 1) which provides cutting-edge technology for obtaining highway pavement information under local conditions. Data collected through the MFNSV is processed through the Hawkeye 2000 package installed on the MFNS vehicle. This study was possible due to the assistance rendered by the Planning Division of the Road Development Authority of Sri Lanka which presently uses the MFNSV for its highway pavement data collection operations. Sample data was collected manually as well as through the MFNSV and results were compared and differences were statistically checked.
According to the manufacturer, the following are some of the general applications of MFNSV [1]:  Network and project level road and asset collection surveys  Routine pavement monitoring surveys  Roadside inventory and asset management  Road geometry and mapping surveys  Contractor quality control  Road safety assessment  Line marking reflectivity  Airport runway inspections Highway pavement information listed below was collected from the sample sites through manual surveys as well as from the MFNSV. 1. Crack areas (all types of cracks) 2. Defected patch areas 3. Ravelling areas 4. Pot-hole areas 5. Lane width Thereafter samples obtained were compared with one another using statistical techniques and the differences were observed.

Methodology
As already mentioned, Manual Data Collection and MFNS Vehicle Data Collection were carried-out at four selected road sections.

Manual Road Inspection
Under this method, one has to inspect the road visually and collect data manually by measuring each and every type of failures and their magnitude in relation to the chainage and fill the standard visual inspection form (see Appendix A) that was developed for this process. Though this form records the potholes in numbers and other failures in linear meters, to compare with the sensitivity of vehicle data, the areas of those failures had to be measured. If a pothole area was larger than 4 m 2 , it was recorded as a surface deformation. The minimum length of a survey section was 100 m. However in special cases, the inspector had to increase the survey length. This is the most accurate method to inspect the road since it is done manually.  This vehicle is of cutting-edge technology, and includes devices such as five numbers of high resolution digital cameras, a front laser mounting beam, side projection lasers, a rotorpulser, a data acquisition system, gipsy-track geometry, a GPS (Global Positioning System) and a DGPS (Differential Global Positioning System). Using those devices, details such as roughness, rutting, geometric data and road condition data can be obtained as images and distance measurements. The Hawkeye processing toolkit is used to process the data taken from this vehicle. To obtain accurate readings, the vehicle has to be driven on the road at a speed between 20 km/h and 110 km/h. Pavement data was collected from four sample stretches each of 200 meter length located on the A004 CRWB(Colombo, Ratnapura, Wellawaya and Batticaloa) road between Meepe and Avissawella. As indicated in Figure  2, four 200 meter stretches from 35 km, 40 km, 45 km and 50 km posts were selected for the data collection.

Failure Types
 Cracks -There are several types of cracks that can be seen on roads. i.e, longitudinal cracks, transverse cracks, diagonal cracks, block cracks and crocodile cracks. All of those cracks can be measured under this category.  Raveling -This is the removal of materials from the pavement surface.  Potholes -Potholes are bowl-shaped holes.
The distress progresses downwards into the lower layers of the pavement.  Defected patch areas where the failure of early treated are as that has taken place.

Measurement of Failures
During manual inspection, all failure areas were measured using steel tapes. In the MFNSV method, all failure areas were measured using the Hawkeye processing toolkit software. The area obtained from the software is presented below.

Data Analysis and Results
Vehicle inspection survey data was collected to find out the total failure percentages according to each failure type with 200m samples taken at every 5km interval by looking at all pictures of the road using the software. All failure areas identified during vehicle inspection were analysed using the Hawkeye software tool. Its calculation part is shown below. All failure areas identified during manual inspection were measured using steel tapes and the corresponding failure percentages were calculated ( Figure 3). Thereafter, the failure percentages were compared and a sensitivity analysis of pavement inspection methods with 200m samples taken at every 5km interval was done. The process followed in obtaining failure percentages of the first section (35 km -35.2 km) is described below in which only the 35.001 km chainage has been considered for the sample calculation.
In manual inspection, Consider the 35.001 km chainage in Table 1 which is shown in APPENDIX B relevant to manual inspection survey results.
Consider the 35.001 km chainage in Table 1 which is shown in Appendix B Band which relates to visual inspection survey results from the vehicle.

Sensitivity Analysis
The vehicle and manual inspection surveys were compared with respect to total failure percentages. Thereafter, the sensitivity of the vehicle was checked through sensitivity analysis. Each 200 m was then divided into 4 m intervals and 50 data sets for manual inspection and 50 data sets for vehicle inspection were obtained for each section. SPSS software was used to check the statistical significance of the test. Results were obtained based on the t-test. For the t-test, both 95% and 99% confidence intervals were used.
Correlation analysis is used to quantify the association between two continuous variables (e.g., between an independent and a dependent variable or between two independent variables). To find the correlation coefficient, the following equation can be used. Regression analysis is a related technique to assess the relationship between an outcome variable and one or more risk factors or confounding variables.
In regression analysis, the dependent variable is denoted by "y" and the independent variables by "x". Many techniques for carrying out a regression analysis have been developed.
Linear regression equation as [2]: Symbol a represents the slope or gradient of the line, and is also sometimes called the regression coefficient while symbol b represents the intercept that is the value of y where the line crosses the y axis.
Null Hypothesis is a statement that the value of the population parameter is no different from that which is equal to a specified value. Statically, a null hypothesis may be stated as; Ho: µ 1 = µ 2 Where µ 1 and µ 2 are population means. The Alternative Hypothesis is a statement that there is some difference in the two populations. It can be stated as Ha : µ 1 ≠ µ 2 For our study, manual inspection readings were taken as µ 1 and vehicle inspection readings were taken as µ 2.
The output taken from the SPSS software is discussed below. It shows that the accuracy of the MFNSV is high. Figure 4 shows the failure percentages of the first section (35 km-35.2 km) which were selected for the study. The X axis represents 50 parts while the Y axis represents failure percentages of each part. Table 1 and 2 represent the tabulations of the mean, standard deviation, standard error mean, confidence interval of the difference, t value, degree of freedom and the significant value of two tailed test for the 95% and 99% confidence intervals respectively.
The paired sample t test results as shown in Table 1 above will appear in the SPSS output window. The p-value is 0.021 and value is 0.05. The p-value is greater than 0.05 which implies that the hypothesis has not been rejected. Therefore, there is no difference between manual and vehicle readings at 95% confidence level.
by "x". Many techniques for carrying out a regression analysis have been developed.
Linear regression equation as [2]: Symbol a represents the slope or gradient of the line, and is also sometimes called the regression coefficient while symbol b represents the intercept that is the value of y where the line crosses the y axis.
Null Hypothesis is a statement that the value of the population parameter is no different from that which is equal to a specified value. Statically, a null hypothesis may be stated as; Ho: µ 1 = µ 2 Where µ 1 and µ 2 are population means. The Alternative Hypothesis is a statement that there is some difference in the two populations. It can be stated as Ha : µ 1 ≠ µ 2 For our study, manual inspection readings were taken as µ 1 and vehicle inspection readings were taken as µ 2.
The output taken from the SPSS software is discussed below. It shows that the accuracy of the MFNSV is high. Figure 4 shows the failure percentages of the first section (35 km-35.2 km) which were selected for the study. The X axis represents 50 parts while the Y axis represents failure percentages of each part. Table 1 and 2 represent the tabulations of the mean, standard deviation, standard error mean, confidence interval of the difference, t value, degree of freedom and the significant value of two tailed test for the 95% and 99% confidence intervals respectively.
The paired sample t test results as shown in Table 1 above will appear in the SPSS output window. The p-value is 0.021 and value is 0.05. The p-value is greater than 0.05 which implies that the hypothesis has not been rejected. Therefore, there is no difference between manual and vehicle readings at 95% confidence level. As mentioned earlier, for a 99% confidence interval the SPSS output window gives the pvalue as 0.021 and value is 0.01. The p-value is greater than 0.01 and this implies that the hypothesis has not been rejected. Therefore, there is no difference between the manual reading and vehicle reading at a 99% confidence interval.   Correlation values obtained for 95% and 99% confidence intervals are presented below.
For both confidence intervals, correlation values are obtained as 0.997. These correlation values are close to +1. Hence it can be predicted that data of both readings are highly related.

Conclusions
It was observed that manual and MFNSV inspection readings were almost the same. However, in some places differences between the two methods were noticed. Manual road inspection was carried-out after the MFNSV data collection. The time duration between the two inspections was around 5 months. It was assumed that there were no significant changes that have taken place in road failures during that time period. However, the gap in the duration in which the two surveys were done will slightly affect the difference in the two readings. For example, the cracks propagating along the road during that time are shown only in the manual inspection. In such places, manual reading takes a higher value than the MFNSV reading and in places where the failures were treated during that time belt the By conducting a sensitivity analysis it was found that the accuracy of the MFNSV inspection is quite high when compared with the manual road inspection. It was evident that, this vehicle can be recommended for inspection of long road stretches, since the manual method takes too much time and since it is labour intensive. Because of the high initial cost, conducting MFNSV inspection on short stretches is uneconomical. For such cases manual inspection is recommended.