Accuracy in mechanistic pavement design consequent upon unbound material testing
As part of a European Union funded research study (the "SCIENCE" project) performed between 1990 and 1993, granular road construction material and subgrade soil specimens were tested in the four participating laboratories of the project: Laboratório Nacional de Engenharia Civil Portugal University of Nottingham United Kingdom Laboratoire Central des Ponts et Chaussées France Delft University of Technology The Netherlands The author was based the first of these and visited the other participating laboratories, performing the majority of the work described. Inaccuracies in repeated load triaxial testing based on the use of different apparatus and instrumentation are identified. A detailed instrumentation comparison is undertaken, which results in the magnitude of potential errors being quantified. The author has derived material parameters and model coefficients for the materials tested using a number of previously published material models. In order to establish these parameters a method for removing outliers from test data based on the difference between the modelled and experimental material parameters for each stress path applied was developed. The consequences of repeatability and reproducibility, variability and inaccuracies in the output of repeated load triaxial testing, on the parameters and, hence, on computed pavement design thicknesses or life is investigated using a number of material models and the South African mechanistic pavement design method. Overall, it is concluded that: • Instrumentation differences are not as critical as variations in results obtained from different specimens tested in a single repeated load triaxial apparatus. It was found that specimen manufacture difference yielded greater variation that instrumentation differences. • Variation in results has some effect on the upper granular layers, where higher stress levels are experienced, but even quite considerable variation in the results from materials used in the lower layers has little effect on pavement life. • Analytical methods to determine the stresses and strains vary considerably as do the predicted pavement thicknesses consequent on using these methods. The inaccuracies in testing (large discrepancies are found when the same material is tested in the same laboratory) and the limitations of the available material models severely limit the usefulness of advanced testing and non-linear modelling in routine pavement design. On the basis of this study it is recommended that a more simplistic pavement design approach be taken keeping in line with future developments of testing and modelling and field validation.