Statistical applications and stochastic analysis for performance-related specification of asphalt quality assurance

Performance-related specification (PRS) is known as the most advanced specification type in asphalt pavement quality assurance (QA). One major feature of PRS QA is the consideration of material variability in performance prediction models through the application of several essential statistical techniques. The uncertainty in material properties can consequently be encoded in the evaluation of asphalt mix quality. For this reason, it is of critical significance to use the statistical applications correctly in PRS QA. One product of the NCHRP 9-22 project is a computerized quality assurance tool, quality-related specification software (QRSS). This computer program can quantitatively evaluate the as-constructed asphalt mix quality against the as-designed mix quality to predict service life. To quantify stochastically the quality of both the as-designed and as-built mixes with the consideration of material variability, several essential statistical techniques were applied to the QA procedure. This paper presents the techniques used and the applications implemented in the QRSS. It includes statistical methods to estimate variance of a multivariate function (e.g., Monte Carlo simulation and Rosenblueth method), conversion methods from distress to service life, expression of predicted distress as well as its service life by using probability distributions, and a method to quantify the relative difference in predicted service life between as-designed and as-constructed mixes with the use of cumulative frequency distributions.