Remaining service life (RSL) has been defined as the anticipated number of years that a pavement will be functionally and structurally acceptable with only routine maintenance. The Kansas Department of Transportation (KDOT) has a comprehensive pavement management system, network optimization system (NOS), which uses the RSL concept. In support of NOS, annual condition surveys are conducted on the state highway system. Currently KDOT uses an empirical equation to compute RSL of flexible pavements based on surface condition and deflection from the last sensor of a falling-weight deflectometer (FWD). Due to limited resources and large size, annual network-level structural data collection at the same rate as the project level is impractical. A rolling-wheel deflectometer (RWD), which measures surface deflections at highway speed, is an alternate and fast method of pavement-deflection testing for network-level data collection. Thus, a model that can calculate RSL in terms of FWD first sensor/center deflection (the only deflection measured by RWD) is desired for NOS.In this study, RWD deflection data was collected under an 18-kip axle load at highway speed on non-Interstate highways in northeast Kansas in July 2006. FWD deflection data, collected with a Dynatest 8000 FWD on the KDOT network from 1998 to 2006, were reduced to mile-long data to match the condition survey data collected annually for NOS. Normalized and temperature-corrected FWD and RWD center deflections and corresponding effective structural numbers (SNeff) were compared. A nonlinear regression procedure in Statistical Analysis Software (SAS) and Solver in Microsoft Excel were used to develop the models in this study.Results showed that FWD and RWD center deflections and corresponding SNeff are statistically similar. Temperature-correction factors have significant influence on these variables.FWD data analysis on the study sections showed that average structural condition of pavements of the KDOT non-Interstate network did not change significantly over the last four years. Thus, network-level deflection data can be collected at four-year intervals when there is no major structural improvement.Results also showed that sigmoimal relationship exists between RSL and center deflection. Sigmoidal RSL models have very good fits and can be used to predict RSL based on center deflection from FWD or RWD. Sigmoidal equivalent fatigue crack-models have also shown good fits, but with some scatter that can be attributed to the nature and quality of the data used to develop these models. FWD data analysis on the study sections showed that average structural condition of pavements of the KDOT non-Interstate network did not change significantly over the last four years. Thus, network-level deflection data can be collected at four-year intervals when there is no major structural improvement.Results also showed that sigmoimal relationship exists between RSL and center deflection. Sigmoidal RSL models have very good fits and can be used to predict RSL based on cen...

Abstract:In the mechanistic-empirical pavement design guide, prediction of flexible pavement response and performance needs an input of dynamic modulus of hot-mix asphalt at all three levels of hierarchical inputs. This study was intended to find the best way to predict/derive this input. Nine Superpave pavement sections were selected as test sections in this study. Deflection data on all test sections was collected with a Dynatest 8000 falling weight deflectometer shortly after construction. The deflection data, normalized with respect to 40-kN load, were used to back-calculate asphalt layer moduli using three back-calculation algorithms. Laboratory dynamic modulus tests were conducted on asphalt concrete ͑AC͒ cores and laboratory-compacted samples. Dynamic modulus was also estimated with the Witczak model, new Witczak model, and Hirsch model. The results show that the AC moduli obtained from various back-calculation programs used in the study are generally comparable. Laboratory dynamic modulus is comparable at 4°C, but the variation increases as the test temperature increases. The Witczak model underestimates the dynamic modulus at low temperature and overestimates it at higher temperature. The parameter estimate when the laboratory dynamic modulus is used as a dependent variable and the moduli from other approaches as independent variables is close to 1. This is especially true for the AC moduli estimated by various prediction methods. The Hirsch model appears to be the best for estimation and is closely followed by the new Witczak model.