Network and Project Level Evaluations Using a TSDD

flexible pavement in Montana

case study summary 

This case study analyzes data collected using a Traffic Speed Deflection Device (TSDD) equipped with Doppler lasers on a flexible pavement highway in the State of Montana. The highway section is divided into uniform segments using the change point detection method. Moving-load backcalculation is then performed for each segment using bakML to obtain the layer moduli. The collected FWD data were unreliable; therefore, the layer moduli could not be verified through nonlinear dynamic backcalculation. Instead, a sensitivity analysis is performed using bakML by vertically shifting the measured slopes, and similar moduli are obtained on average for the different scenarios. Overlay design is then performed using the classical mechanistic–empirical (ME) method by considering the backcalculated layer moduli, seasonal variations, and actual truck traffic. The analysis results in a minimum required overlay thickness of 2 inches. This type of structural evaluation can be implemented at both the network and project levels by adjusting the number of segments and the level of detail in the analysis. Furthermore, network-level performance metrics are presented, including Surface Cracking Life (SCL) and Pavement Deformation Life (PDL), which estimate the remaining service life (in years) for surface cracking and deformation of the unbound layers, respectively.

case study insights

The highway section included in this study is part of US Route 2 (US-2), a two-lane road located west of Chester, Montana. It extends from reference post (RP) 308.0 to RP 321.4, covering a length of 13.4 miles. The lanes are 11.8 ft wide, and have shoulders on both sides with variable widths along the highway. 


The pavement structure is composed of a three-layer flexible system consisting of an asphalt concrete layer and an aggregate base layer over a subgrade layer. Ground Penetrating Radar (GPR) testing was performed in the eastbound direction on August 21, 2023 using two antennas, having frequencies of 2 GHz for the air-coupled antenna and 400 MHz for the ground-coupled antenna. The thicknesses are shown to be variable, which is typical for a highway construction, with an average of 4.5-inch for the asphalt concrete layer and 20-inch for the aggregate base layer. 


As part of the pooled fund study, TSDD testing is performed on August 10, 2024, in the afternoon using ARRB’s iPAVe system, which uses the Greenwood Traffic Speed Deflectometer (TSD). The data are reported at 0.01-mile (52.8 ft) intervals.

highway segmentation

Change point detection is applied to subdivide the highway in the eastbound direction into uniform segments using multiple variables, including TSDD deflection basin indices (SCI, BDI, and BCI) and pavement layer thicknesses (hAC for asphalt concrete and hAB for aggregate base). The analysis identifies 13 segments, with lengths ranging from 0.25 to 3 miles. The variables are normalized in the plot below for visualization purposes. The segmentation can be refined by adjusting the penalty value to control the number of segments.

moving load backcalculation and mechanistic-empirical design

Backcalculation of the TSDD data is performed for each uniform segment using bakML, and the required overlay thicknesses are determined using paveSiME, a pulsuus application that employs the classical mechanistic–empirical (ME) design method for pavement structural evaluations.


The asphalt concrete is modeled as a linear viscoelastic material in bakML, while the unbound layers are modeled as linear elastic. A review of the backcalculated layer moduli from the TSDD indicates that the subgrade is relatively stiff, suggesting possible stabilization, with moduli ranging from 14 to 26 ksi. In contrast, the base layer moduli are typical of a crushed aggregate base, with values ranging from 23 to 43 ksi. Finally, the asphalt concrete layer moduli at 77°F are low, likely due to damage and the use of a softer binder for cold climates.


Overlay design is performed by accounting for seasonal variations, including freezing and thawing of the unbound layers in this cold environment, along with the actual traffic over a 20-year design period. The analysis considers fatigue cracking of the asphalt concrete layer and permanent deformation of the base and subgrade layers. Deterministic sampling is used to account for variability in layer thicknesses and layer moduli within each segment.


The figure shows the calculated overlay requirements, which are typically reported as the 84th percentile (average plus one standard deviation) after correcting areas of localized failure (R&R). This results in a 2-inch overlay, which is the minimum structural thickness for highways (3 inches for airports).


An overlay thickness is typically acceptable for network-level evaluation. However, different alternatives can be recommended at the project level, depending on pavement condition, the critical layer (i.e., the first layer to fail), and functional classification. These alternatives include mill and overlay, recycling, and other treatments. A project-level evaluation will require shorter segments and accurately determined pavement layer information.

bakML sensitivity analysis

FWD testing was performed on August 22, 2023, in the morning using a truck-mounted deflectometer at 100 m intervals in the eastbound direction. Initial analysis of the load and deflection time histories indicated that the FWD data is unreliable; therefore, the data cannot be used to verify the results from bakML using nonlinear dynamic backcalculation of FWD data. Montana Department of Transportation (MDT) has since acquired a reliable FWD, along with a 3D GPR system, for the structural evaluation of the large network owned and maintained by the agency.


Due to the absence of reliable FWD data, a sensitivity analysis is performed using bakML by varying the TSDD-reported slopes by ±10 microns to quantify the effect of potential errors in the TSDD zero-slope assumption. The bar chart shows that the layer moduli for the three cases (control and ±10 microns) are comparable, with the largest difference being about 10% for the subgrade layer.


Regardless of this analysis, FWD testing using a reliable device and nonlinear dynamic backcalculation are required to verify the accuracy of the TSDD slopes and the resulting layer moduli.

network-level metrics

pulsuus developed the Surface Cracking Life (SCL) and Pavement Deformation Life (PDL), network-level metrics determined from the TSDD or FWD for estimating the remaining life in years for surface cracking and unbound layers deformation, respectively. The SCL, shown with high-severity cracking, aligns well with observed distress. Segment 5, with a low SCL of 1.5, is the next likely to fail, consistent with its lowest base modulus of 23 ksi. In contrast, segment 2 with its highest base and subgrade moduli has the lowest amount of cracking with only 9 ft per mile, and has one of the highest SCLs. The PDL is not plotted since the subgrade was not found to be the critical layer for this highway. Together, SCL and PDL help prioritize segments for project-level evaluation when paired with distress data.

acknowledgments and disclaimer

pulsuus acknowledges the Montana Department of Transportation (MDT) for providing access to the TSDD data and the pavement layer thicknesses used in this case study. Furthermore, MDT assumes no responsibility for any recommendations or conclusions derived from this publicly available information.