Efficient operations of highways are inevitably critical for the sustainable, fast and safe transportation of people and freight. Effective maintenance in particular is of primary importance for ensuring that a highway meets operational and safety standards and offers efficient services to its users. Road surface or pavement maintenance comprises of required actions for planning rehabilitation strategies and maintenance activities over a highway network, having as an objective to keep the network in an acceptable functional and structural condition. Therefore, pavement condition assessment and prediction of deterioration become mandatory prerequisites for deciding on maintenance programming and budget allocation. Improved models of that category would better decisions on maintenance activities and allocation of resources. In this context, this PhD research focuses on the development of novel, efficient models for assessing existing condition and predicting future damage of pavements. Particular research efforts include introduction of an approach for assessing existing pavement condition, as well as the development of new post-treatment pavement deterioration models. New tools and methods are developed for that purpose: first, a new method for assessing existing pavement structural health, using the dynamic stiffness modulus is introduced and evaluated. Next, a Bayesian duration based model is developed in an effort to predict remaining service life of pavements, following rehabilitation actions. A subsequent model is developed for forecasting asphalt cracking initiation, again following maintenance activities; this model is based on Artificial Neural networks. Explanatory parameters are identified in all models; these include treatment activities, weather and structural characteristics of pavements. A qualitative comparison of BDM and ANN techniques reveals conceptual differences: BDM refers to the temporal behavior of pavement deterioration while attempts to forecast the probability of deterioration given a set of inputs. On the other hand, the quantitative comparison yields strong statistical similarities between the outcomes of the two models, with respect to estimating post treatment deterioration timing.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:723399 |
| Date | January 2016 |
| Creators | Karlaftis, Aristeidis G. |
| Publisher | University of Bolton |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://ubir.bolton.ac.uk/1197/ |
Page generated in 0.0022 seconds