A calibrated residual pavement life determination for unbound pavements

Abstract

Over the next 3 years Waka Kotahi NZ Transport Agency (NZTA) is planning to invest $2.8 billion dollars in maintaining and operating the state highway network and $2.6 billion improving state highway assets, which are currently valued at approximately $52 billion. A further $4.2 billion is planned to support local government in the operation, maintenance and renewal of the local road network. Given the preceding levels of investment, determining the as-built life of new unbound granular pavements is becoming increasingly important for two reasons. The first is the increasing use of “design and construct” type contracts where the client is required to make some judgement on the length of life of the finished pavement at the end of the defects liability period before accepting the project. Second, it is imperative to ensure that all designs are as efficient and effective as possible to minimise maintenance expenditure and increase sustainability.

Consequently, the aim of this research was to develop and calibrate a residual pavement life model capable of improving upon current estimates of remaining life for newly constructed unbound granular pavements. Specific objectives of the research included: • Clearly identify the end of life conditions. Currently, the “end of life” is an ill-defined mix of functional requirements and maintenance costs; • Develop improved pavement response and performance models for unbound pavements; • Use advanced laboratory testing concepts, typically used for pavement design, to improve residual life prediction; • Incorporate quality assurance data to improve residual life prediction; and • Use early life performance data to improve residual life prediction. In order to determine end of life conditions, a two-stage investigation of the NZTA’s Long Term Pavement Performance (LTPP) database was undertaken. The first stage was a Kaplan Meier Survival analysis of rutting data. Cox proportional hazards models were used as a second stage to investigate other distress modes and determine their influence on rehabilitation decisions. The analysis concluded that the median survival estimate of the upper quartile rutting statistic was a suitable end of life condition to use.

To develop improved pavement response models for unbound pavements, relatively simple non-linear response models for axi-symmetric finite element analysis were examined. The non-linear models provided reasonable estimates of elastic strain data measured during accelerated pavement testing (APT). The non-linear models were a significant improvement on the linear elastic models currently used in residual life calculations.

A review of existing performance models found that the form proposed by Wolf and Visser (1994) was the best fit to the observed APT performance. In creating a new Pavement Layer Performance Model, it was also important to recognise limitations in the proposed response and performance models and include the concept of shake-down that had been noted in the literature.

To use advanced laboratory testing concepts to improve life prediction, the literature was reviewed and found to contain several promising performance models for Repeated Load Triaxial (RLT) testing. However, none matched the mathematical form of the deterioration observed in the APT testing. A new performance model was therefore developed from existing laboratory RLT data of materials used in the APT testing. The new RLT Performance Model predicted permanent strain at varying levels of resilient strain and was incorporated into the Pavement Layer Performance Model. The new model matches the behaviour seen in APT pavements and statistical analysis concluded that the RLT Performance Model did improve the prediction of residual life.

The fourth objective was to incorporate Quality Assurance data into the model to improve life prediction. The literature suggested that including Quality Assurance data would improve the prediction. The analysis undertaken concluded that adding quality assurance data, such as density, was not needed. The FWD data used for the non-linear response model and the new Pavement Layer Performance Models had already accounted for changes in quality. The final objective was to use early life data to improve life prediction, as the literature review found that advanced pavement design models had difficulty predicting early life performance. Including early life performance data in the residual life model, typical of that available at the end of the defects liability period, compensated for the difficulties and improved the life estimates.

In summary, the overall aim of the research was to develop a calibrated residual pavement life model for unbound granular pavements that was an improvement on existing approaches. The new model was compared against the best existing model, with the new model providing a significantly improved estimate of residual life. The new model can be used to provide project level assessments of the remaining life of newly constructed unbound granular pavements. However, a period of field validation is recommended, and the further research recommended is likely to significantly enhance the model’s accuracy and applicability to a wider range of scenarios.