U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000


Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

 
REPORT
This report is an archived publication and may contain dated technical, contact, and link information
Back to Publication List        
Publication Number:  FHWA-HRT-12-023    Date:  December 2012
Publication Number: FHWA-HRT-12-023
Date: December 2012

 

Simplified Techniques for Evaluation and Interpretation of Pavement Deflections for Network-Level Analysis

CHAPTER 6 - Guidelines for Consideration of Time of Day and Season of Year for Optimal FWD Data Collection for PMS Applications

Summary of Approaches from Literature Review

Since little has been published until recently about including deflection data in PMS applications, it stands to reason that even less has been published on the subjects of time of day and season of year when FWD testing should take place.

However, in July 2009, the Michigan Department of Transportation (MDOT) mentioned the potential of utilizing FWD load deflection testing at the network level.(30) According to the report, MDOT intended to prescribe 500-ft (152.5-m) testing intervals, or about 10 test points per mile of pavement. MDOT also mentioned spring and late summer/early fall; however, no particular time of day to conduct deflection tests was mentioned.

Other agencies by and large either do not utilize load deflection tests at the network level, or the few that already use deflection testing generally allow their engineers to fix the proper test point intervals and/or the time of year to test, depending on the application and climatic region. For example, AkDOT also utilizes 500-ft (152.5-m) test intervals; however, it does not yet use deflection testing on their entire roadway network and is currently adding FWD test results as time allows (only the Parks Highway has been completed to date). The viable testing season is vastly shortened in Alaska due many months of frost penetration as well as the occasional year-round permafrost presence.

The vast majority of deflection testing-whether at the network or project level-is generally carried out during normal daytime work hours. Similarly, with seasonal testing, it is generally believed that the gathering of deflection data should occur when these data can be gathered as climatic conditions allow (and the pavement temperature is above freezing). This is to allow agencies to carry out network-level testing as far as possible at their own convenience because what is convenient as far as timing goes will most likely produce good deflection data for use in the stochastic relationships developed for this project.

While there are exceptions to the above statements, particularly when concrete pavements are tested, the general rule from the literature and practice alike is to test whenever possible during the day or year and not restrict when an agency can test for PMS applications. At some point in the future, it is also likely that the RWD will become a feasible and reasonably low-cost alternative to the FWD; but this is not currently the case, so this eventuality is not considered in this report.

It is also important to consider the issue of protection and maintenance of traffic (P&MT). For most medium to high traffic roads, FWD testing must be fit around the periods of high traffic flow, which often happens in the morning (6 to 9 a.m.) and afternoon periods (3 to 5 p.m.).

RESULTS OF LTPP DATA ANALYSIS

The LTPP database was reviewed in detail to ascertain whether limitations should be imposed, or suggested, to either the time of day or season of year when network-level FWD testing should (or should not) take place. At the outset, it was assumed that any significant limitations of this nature would tend to discourage the use of simplified techniques for both gathering and analyzing load deflection data, since roadway networks are, by definition, very extensive compared to individual projects.

DEFLECTION ADJUSTMENT PROCEDURES

The main reasons for adjusting the measured pavement deflections (at a given load level) is pavement temperature for flexible pavements or temperature gradients for PCC pavements.

For flexible pavements, in general, the use of an adjusted deflection level due to pavement temperature was not observed. Figure 62 shows the accuracy of the fatigue cracking model developed in chapter 4 for flexible pavements. The figure shows the ROC curve for two deflections, one adjusted by temperature and the other not adjusted. The fatigue cracking model indicates no improvement whatsoever by adjusting D1 due to pavement surface temperature variation. A similar analysis is provided in figure 63 for rutting. The statistics of these regressions are shown in table 37 and table 38.

Figure 62. Graph. Specificity of fatigue cracking models for flexible pavements. This graph shows a scatter plot of specificity of fatigue cracking models for flexible pavements. The x-axis represents 1 minus specificity, and the y-axis represents sensitivity. Both axes range from zero to 1. There are four data series in this plot. The first is the reference line (shown as a lightly shaded line), the second is the unadjusted deflection line (shown as an arched solid line), the third is the adjusted deflection line (shown as an arched dotted line), and the fourth is the idealized model (shown as a dashed line). The reference line is a straight diagonal line that starts at coordinates 0,0 and ends at coordinates 1,1. Both the adjusted and unadjusted deflection lines start and end with the same coordinates as the reference line, with intermediate points located to the left and above the reference line. The area under both curves is the same because these curves follow almost the same trend and equals 0.64. The idealized model starts at the origin, continues vertically to the point with coordinates 0,1, and then continues horizontally to the point with coordinates 1,1. The area under the idealized model is equal to 1.
Figure 62. Graph. Specificity of fatigue cracking models for flexible pavements.

Figure 63. Graph. Specificity of rutting models for flexible pavements. This graph shows a scatter plot of specificity of rutting models for flexible pavements. The x-axis represents 1 minus specificity, and the y-axis represents the sensitivity. Both axes range from zero to 1. There are four data series in this plot. The first is the reference line (shown as a lightly shaded line), the second is the unadjusted deflection line (shown as an arched solid line), the third is the adjusted deflection line (shown as an arched dotted line), and the fourth is the idealized model (shown by a dashed line). The reference line is a straight diagonal line that starts at coordinates 0,0 and ends at coordinates 1,1. Both the adjusted and unadjusted deflection lines start and end with the same coordinates as the reference line, with intermediate points located to the left and above the reference line. The area under the unadjusted deflection curve is 0.67, and the area under the adjusted deflection curve is 0.62. The idealized model starts at the origin, continues vertically to the point with coordinates 0,1, and then continues horizontally to the point with coordinates 1,1. The area under the idealized model is equal to 1.
Figure 63. Graph. Specificity of rutting models for flexible pavements.

 

Table 37. Statistics for flexible pavement fatigue cracking models: D1 versus D1 adjusted.

Model Significance Area Under
ROC Curve
Correct Cases- Modeling percent)

D1

0.53

0.64

59.0

D1adj

0.73

0.64

61.7

 

Table 38. Statistics for flexible pavement rutting models: D1 versus D1 adjusted.

Model Significance Area Under
ROC Curve
Correct Cases-Modeling(percent)

D1

0.40

0.67

68.3

D1adj

0.12

0.62

59.5

While the fatigue cracking model showed virtually no change in either the area under the ROC curve or the correct cases predicted by the model, no improvement was noted. For the rutting model, on the other hand, adjusting D1 made the model inferior to the unadjusted model.

In the case of the roughness model, deflection parameters involving D1 were not among those that showed the most promise overall. Therefore, the question of adjusting D1 or not is mute.

It was concluded that an adjustment to D1 due to asphalt temperature variations is neither helpful nor worth the time and effort to carry out this extra step for applications at the network level. Such an adjustment would also cause agencies to have to gather even more data-in this case, air and pavement surface temperature plus the previous day's high and low temperatures at a minimum. To keep the network-level PMS application of adding simplified deflection analysis techniques to routine pavement management data would be counterproductive and is ill advised.

Since concrete surfaces are not subject to the same visco-elasticity that asphalt surfaces are, a temperature correction for concrete surfaces was not investigated.

Recommendation: Do not adjust deflection measurements – not even D1 – due to temperature variations in any visco-elastic surface course layer.

Recommendations for Time of Day for Deflection Measurements at Network Level

As mentioned in the previous section, "Results of LTPP Data Analysis," no limits for agencies are recommended for the time of day for FWD load deflection testing except normal daytime working hours. The main reason for this is to encourage agencies from doing network-level deflection testing by restricting the time of day when measurements should be conducted. A secondary but equally important reason is that the time of day generally does not appreciably affect the accuracy of the simple models used by the logistic model, as recommended in this report. Networks are very large compared to individual projects, and since at least a statistically significant sampling of each uniform or standard pavement management section should be sampled, it would indeed be discouraging for an agency to also have to further limit their FWD testing window due to the large number of lane miles involved to cover an entire network.

Exceptions to this rule may apply to jointed PCC pavements when large temperature gradients within the slab are present. Another possible exception is very high volume urban freeways when traffic control will be an issue. In such cases, only late evening through early morning testing is feasible and should be carried out in lieu of doing nothing. Finally, running FWD load deflection tests while the underlying unbound layers may be frozen, or during the spring thaw in many of the northerly regions, should also be avoided for network-level FWD testing.

Exactly when concrete pavements have excessive temperature gradients and thus curling or warping due to these gradients is largely a function of local weather patterns, temperature differentials between day and night air temperatures (e.g., in desert regions), and when there is little or no cloud cover during the daytime, in which case, there is probably not a significant temperature gradient. If highly variable daily temperature swings occur on a 24-h basis, it is generally recommended to only test during morning hours before the sun has excessively heated the concrete slabs at the top (thus causing slab warping).

Recommendation: Limit the agency's time window for testing flexible pavements to normal daytime working hours. Oftentimes, jointed concrete pavements are an exception to this rule, depending on climatic zone and/or temperature gradients in the slab, etc. Urban freeways are also a potential exception for safety reasons. Finally, do not test when the unbound materials are frozen beneath the pavement or during spring thaw.

Recommendations for Season of Year for Measurements at Network Level

An analysis and review of the FWD data in the LTPP database reveals that testing at virtually any (completely) thawed time of year is acceptable for network-level deflection testing. An example of the LTPP data analysis is shown in table 39.

Table 39. Statistics for flexible pavement IRI as a function of season.

Statistics All Data Spring Summer Fall Winter

Accuracy of model

0.73

0.65

0.78

0.77

0.80

Correct prediction rate (percent)

68

56

68

69

77

Error type I (false positive) (percent)

6.6

5.3

2.5

5.1

7.5

TPR (percent)

69

54

63

68

81

TNR (percent)

64

68

88

71

67

Note: Error type I predicts acceptable performance incorrectly, TPR predicts correctly acceptable performance, and TNR predicts correctly not acceptable performance.

In this example and almost all other seasonal examples, there is little important difference between the overall goodness-of-fit and the corresponding values where the data are broken down by season. In table 39, the apparent improvement for some statistical categories during winter testing is due to the fact that only two of the four climatic regions are represented-DNF and WNF. Error type I (false positive), the most serious of the two potential error types, is evidently improved by using summer season data only, which is the time of year that is most likely suitable for field testing and scheduling for most agencies.

For network data collection purposes, deflection testing during the spring thaw period for the DF and WF zones should be avoided because these data tend to cause bias results toward rehabilitation of those pavements that were tested during spring thaw.

Recommendation: Do not limit an agency's seasonal time window for network-level deflection testing except during spring thaw conditions.

Applications and Limitations

The real issues indicted in the foregoing sections are that depending on location and environmental circumstances, the overall recommendations as to when to conduct FWD test surveys at the network-level may differ significantly. Obviously, limitations will inhibit any potential network testing program to the degree these limitations impose testing restrictions not generally encountered otherwise by agencies that manage roadway networks.

In areas such as Florida and coastal California, winter restrictions are more of an oxymoron than a real test scheduling hindrance, so winter testing can and should take place. Conversely, in Alaska, the testing season is greatly shortened for equally obvious reasons. Finally, the deciding factor for mainline highways, especially urban freeways, may be traffic control. Accordingly, sometimes testing with the FWD will have to take place during nighttime closure hours-if it can take place at all.

Regard the above suggestions and recommendations as being general in nature, not specific for every region of the United States and Canada that was covered by the extensive LTPP monitoring and testing program, which was started by SHRP in the late 1980s.

 

Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101