MEPDG Traffic Loading Defaults Derived From Traffic Pooled Fund Study
CHAPTER 2 - Review of MEPDG traffic loading defaults
This chapter covers the following topics:
- A review of the original MEPDG NALS defaults and the methodology used to estimate or determine global defaults.
- An assessment of the applicability and limitations of the current
MEPDG traffic defaults.
- Recommendations for enhancement of MEPDG traffic loading
defaults.
ASSESSMENT OF METHODOLOGY USED FOR ORIGINAL TRAFFIC
LOADING DEFAULTS
The research team reviewed the original methodology for
generating MEPDG traffic loading defaults, and the results of that review are
summarized in this section.
Data Selection Criteria for Development of MEPDG Defaults
The researchers used the following data selection
criteria to identify data for the development of the original MEPDG traffic
defaults:
- Availability of at least 210 days of AVC data to develop truck
volume-based defaults.
- Availability of at least 1 weekday and 1 weekend of WIM data per
quarter (preferably at least 1 week per quarter) to develop axle loading
defaults.
- Availability of above data items for at least 2 years in a 5-year period.
Development of the default axle load spectra was based on
data from 134 sites. Defaults for axle spacing in tandem and tridem axle
configurations were based on data from 26 sites, and axles per truck type
defaults were based on data from 16 sites. All the defaults were based on LTPP
data collected up to 1999 for General Pavement Studies (GPS) sites that passed
rudimentary LTPP quality checks.
Summary of Traffic Loading Defaults and Methods Used
WIM data were used to generate the following global traffic loading defaults when
sufficient traffic weight data were unavailable:
-
One set of default NALS for each axle group type (single, tandem,
tridem, and quad axles) and FHWA vehicle classes 4 through 13, as applicable. The
default NALS included in the MEPDG were determined based on the following
points and hypotheses:
- Functional
classification was initially used to segregate the LTPP sites for computing
NALS. It was found that NALS were independent of functional class, so one set
of values was determined for NALS for each axle group type and truck class.
- The site-specific
NALS or the annualized axle load spectra were based on WIM data.
- Researchers
calculated a representative NALS for each site that satisfied sufficient data
criteria and passed rudimentary LTPP quality checks and then averaged NALS
across similar sites to develop the global NALS default.
- Assumptions
when weight data were not available to generate the traffic loading defaults
include the following:
-
NALS by axle
group type for each vehicle class remain constant from year-to-year unless
there are regulatory or economical changes that affect the maximum axle or
gross vehicle loads.
-
NALS by axle
group type and vehicle class do not change throughout the time of day or over
the week (weekday versus weekend and night versus day).
-
NALS for each
axle group type and vehicle class do not change from site-to-site within a
specific region or roadway functional classification.
-
One set of typical axle spacing for each of the FHWA vehicle
classes 4 through 13
(i.e., tandem and tridem axle configurations).
- Based on per
vehicle counts from WIM data.
- Comparison of
truck industry values to the values calculated from WIM data.
- Where axle spacing
does not change over time for a site or roadway.
- When there are no
quad defaults.
-
One set of axles per truck type (i.e., single, tandem, and
tridem) for each of the FHWA vehicle classes 4 through 13.
- Based on
spacing reported in per vehicle records (PVRs) obtained from WIM data.
- Where number of
axles for each axle group type was reviewed from the individual truck record
data for a sample of sites.
- Where numbers
of each axle group type were summed for each vehicle class. The total number of
each axle group type (i.e., single, tandem, and tridem) was divided by the
total number of trucks/vehicles to determine the average number of axles for
each axle group type for each truck/vehicle class.
- The number of
axle group types per truck class does not change over time for a site or roadway.
- When there are
no quad defaults.
Determination of Expected Errors in Traffic Estimates Using Original Methodology
The NCHRP Project 1-37A report includes a procedure to
estimate the expected error of the site-specific traffic estimates based on the
amount of data collected at a site, given the variation in the data and the
selected confidence interval.(3) The following equation was used to
calculate the expected error in estimating the daily number of trucks for each
vehicle class.(6)
Figure 1. Equation. Expected error in estimating the daily number of trucks for
each vehicle class.
Where:
e(VCk)j = Expected error for vehicle class k in season j.
Z = Confidence interval coefficient.
n = Number of sampling days.
σ = Standard deviation of the number of class k vehicles in the population during season j.
μ = Mean number of class k vehicles in the truck traffic population during season j.
This approach accounts for errors in truck volume estimates associated with data availability and variability of data due to natural fluctuations in truck volumes during the season. However, it does not account for the errors associated with WIM equipment performance or expected level of accuracy of WIM systems, nor does it provide an estimate of expected errors in axle loads.
APPLICABILITY OF THE ORIGINAL MEPDG TRAFFIC LOADING DEFAULTS
The research team identified, reviewed, and compiled studies published prior to 2009 that evaluated the reasonableness of the original MEPDG traffic defaults or MEPDG sensitivity to traffic inputs. (See references 7-10.) Most of the identified studies focus on sensitivity of pavement design to MEPDG defaults (level 3 inputs) versus site-specific or regional traffic data (level 1 inputs) and present State-sponsored research projects with the focus on State implementation of the MEPDG. Depending on pavement type, distress type, climatic region, State, and functional type of road, the effect of load spectra on pavement design or performance prediction ranged from low to high.
Understanding the Physical Meaning of the Original Default NALS
Some of the criticism regarding the original MEPDG traffic loading defaults comes from unusual shapes reported as a default NALS. The following example demonstrates that the use of the average NALS for each axle group type for a specific truck class is the reason why some NALS do not exhibit some of the typical loading features or patterns that have been reported for specific roadway segments (i.e., appearing like a "stretched" distribution with
longer tail of the distribution (higher percentages of overloads and higher percentage of light loads)).
Figure 2 compares the default NALS for tandem axle for truck class 9 to the loading patterns from three loading spectra for SPS TPF sites within the same functional classification (rural other principal arterial (ROPA)). As shown, site-specific NALS are significantly different, and one may question the applicability or adequacy of the default tandem spectra.
Figure 2. Graph. Comparison of three tandem axle loading patterns to the MEPDG default NALS for tandem axles for truck class 9.
The NCHRP 1-37A research team computed an average NALS for each axle group type and truck class using data from all available sites.(3) The averaging method is used in figure 3, where three different SPS TPF NALS for the same road functional class are averaged and compared to the default NALS for tandem axles. The average of the three site-specific distributions has bi-modal distribution and shows more resemblance to MEPDG defaults.
Figure 3. Graph. Average tandem axle loading patterns for truck class 9.
In summary, the MEPDG default values represent an average NALS from multiple sites across the United States,
including light, bi-modal, and heavy loading patterns. These defaults are based on data collected using a variety of data collection equipment with various maintenance and calibration procedures applied. Thus, the bi-modal distribution
for tandem axles shown in figure 3 is somewhat smoothed out because of the averaging process. The averaging process was used because the amount of variation within the same functional classification found during NCHRP 1-37A was about the same as between all functional classifications, and only one global set of defaults was developed. However, the question still remains-are these overloads real or a result of measurement errors or improper calibration of the WIM equipment? This question was not addressed during NCHRP Project 1-37A; it was assumed that the overloads were
real values.(3)
Applicability of Default NALS
The current state of knowledge does not provide a conclusive answer whether the original defaults could be used successfully by States that move forward with MEPDG implementation. Early findings suggest that while some States find the NALS defaults applicable for their local conditions, others find a need for regional defaults or some other groupings of NALS for improving on the accuracy of pavement design. States that have percentages of heavy and overloaded axles (over 75 percent of legal weight limit) similar to the ones used in default NALS are more likely to benefit from the defaults than States that have significantly different percentages of heavy and overloaded axle loads.
Applicability of Axles per Truck Numbers and Axle Spacing
The values computed under the NCHRP 1-37A study and included in MEPDG software were based on data from 16 LTPP sites that did not have well-documented vehicle classification algorithms.(3) Additionally,
some differences existed in State-defined vehicle classification schemes. Therefore,
it is expected that some inconsistencies or truck misclassifications are present
in the data used to determine the average number of axles per truck type by
axle group and the average axle spacing.
CONCERNS AND LIMITATIONS OF THE ORIGINAL MEPDG TRAFFIC DEFAULTS
The following lists provide a summary of concerns andissues found in application of the original MEPDG traffic defaults.
Data processing and data quality concerns include the following:
- The quality of WIM data was not well known at the time of this study.
The extent and results of WIM equipment validation and calibration activities
was not documented partly because LTPP traffic data collection and processing were
shared between participating States and FHWA. WIM equipment type, performance
requirements, calibration protocols, and data collection algorithms varied from
State to State.
-
Data collected by highway agencies were not of uniform or defined
quality. For many sites, it was impossible to determine if WIM scales were
calibrated and what the outcomes of calibration procedures were.
-
Although not a concern, a constraint of the analysis was that the
data format and processing time made the task of data assessment and defaults
development time consuming and labor intensive.
-
Inclusion of sites with very low axle counts in underrepresented
vehicle classes resulted in choppy distributions that do not follow expected
shapes. This applies to vehicle classes 6-8 and 10-13.
Issues associated with the methodology include the following:
- Only one set of global default NALS exists for a given vehicle
class, limiting pavement designers' ability to analyze different pavement
loading conditions and the effect of these different loading conditions on
pavement response and performance.
- Loading data were collected using different vehicle
classification algorithms, and the consequences of aggregation of the data
collected from the States utilizing multiple vehicle classification algorithms
were not investigated.
- Applicability of the defaults for the States that are using
different classification schemes was not investigated.
- The data quality was unknown, and the NCHRP 1-37A panel
requirement to determine the default NALS values using roadway functional
classification caused high variability and challenges in identifying different
loading patterns.
- The data were not selected randomly and may not represent typical
traffic loading conditions observed on U.S. roads.
- Very few of the sites used to determine the truck default values,
especially NALS, were located on minor arterial and collector highways and on
urban highways, which limits the application of the global NALS.
- Sensitivity of pavement performance prediction models to
variability in NALS was estimated based on using the equivalent annual modulus
concept for rutting and fatigue or alligator cracking (bottom-up cracking) for
new flexible pavements because the pavement performance prediction models or
transfer functions based on the incremental damage concept were not yet
programmed at the time traffic defaults were generated.
RECOMMENDATIONS FOR ENHANCEMENT OF THE ORIGINAL MEPDG TRAFFIC LOADING DEFAULTS
Several recommendations for enhancements or alternatives for the default NALS are as follows:
- Use of new SPS TPF WIM data to develop alternate NALS defaults:One of the limitations associated with the original NALS defaults was lack
of knowledge of WIM data quality and the multitude of different data collection
practices implemented by States that supplied WIM data to LTPP. Utilization of data from SPS TPF WIM
sites
will have a major benefit with respect to known data quality and uniformity in
vehicle classification algorithm. However, data are available for only 26 WIM
sites located in
22 States, so the new NALS defaults may not be applicable to all loading
conditions across the United States.
- Use of loading patterns in defining new default NALS: An improvement in the application of NALS default values within the MEPDG
would be the analyst's ability to specify how the loading patterns can vary in
terms of loads and truck volumes by type
of truck for a given design scenario. While a default or typical distribution
should be
used if the loading condition for a particular pavement design location is
unknown, the availability of alternative loading conditions is highly
beneficial for "what if" sensitivity analyses and for sites that are expected
to have either heavier or lighter than default loading conditions. SPS TPF data
should be used to investigate if different axle loading patterns can be
identified. These patterns should be assessed using the MEPDG to quantify the
significance of pavement design and analysis outcomes for different NALS from
these 26 sites. Based on the results of the analysis, NALS representing
different axle loading conditions should be developed for different vehicle
classes and axle group types.
- Focus on heavy loads and MEPDG predictions: The amount of heavy and overloaded axles drives pavement design decisions. Therefore, heavier load bins should be used in judging whether NALS are different or
statistically the same for pavement design purposes. Similarly, greater emphasis should be placed on the accuracy of data for the heavier load bins. The MEPDG can be used to determine and compare the difference in predicted distress and/or expected service life. Based on the results, the sites can be grouped into those that result in similar predictions and those that are different.
- Comparison of MEPDG outcomes: Both the original defaults and the alternate defaults should be used in the MEPDG analysis to quantify the difference in pavement design outcomes using the NCHRP 1-37A and
SPS TPF default NALS.(3,2)
- Variability in NALS for the road functional classes:In the NCHRP 1-37A study, large differences of the NALS within the same functional classification were observed.(3) There was as much
variation in the normalized values of the heavier load intervals within the same functional classification as between all functional classifications. This can be attributed partly to the limited amount of high-quality traffic data used in that analysis. It would be important to evaluate if the same high loading variability is observed
for the loading spectra obtained for sites within the same road functional class.
- Axle per class (APC) coefficients using SPS TPF data: The number of axles per vehicle class should be determined for the 26 SPS TPF sites, compared to the default values, and used to develop new default values, if needed. This would indicate whether the LTPP vehicle classification algorithm leads to different APC numbers.
- Axle spacing using SPS TPF loading data: Average axle spacing or wheelbase information is used for MEPDG applications involving top-down slab cracking failure in jointed plain concrete pavements (JPCPs). The MEPDG considers wheelbase of the tractor unit on vehicles in classes 8 through 13 in the form of three
inputs: percentage of tractor units with short, medium, and long wheelbases. By default, the MEPDG software assumes an even distribution of short, medium, and long axle spacing occurrences at 33, 33, and 34 percent, respectively. Axle spacing and tractor wheelbase information should be obtained from per-vehicle
records for SPS TPF sites and analyzed to provide updates to the default values. In addition, axle spacing for tandem, tridem, and quad axle groups should be computed and compared with the existing MEPDG default values.