Assessment of Hydraulic and Structural Behavior of Inundated Pavements

Severe flooding events have increasingly inundated the transportation infrastructure, including road networks, across the world and the Unites States. Such extreme events induce significant ingress of water into the pavement layers, which in turn leads to structural defects due to the reduction of the shear strength and resilient moduli of the soil layers, asphalt stripping in flexible pavements, pumping effect in rigid pavements, migration of fines into drainable layers, and swelling of expansive soils. Consequently, premature pavement failure may take place that can impose unforeseen maintenance/rehabilitation costs to highway agencies. Although some experimental research studies have been conducted to assess the impact of flooding on pavement structures, very few published records on the theoretical/mechanistic studies exist in this field. Since the scope of experimental works are usually limited, it is critical to develop a more general mechanistic framework for the assessment of various inundated sections. This study aims at developing a framework for rapid and reliable prediction of the structural response of flexible pavements to major rainfall events and flooding. After reviewing the literature, relevant software packages were selected based on their relative merits of public availability and the accuracy of their results. The selected software packages were utilized for hydraulic and structural modeling of inundated pavement sections. Constitutive models and empirical correlations were extracted from the literature to achieve optimal accuracy of the models with minimum number of inputs. The developed models were then validated with experimental measurements and the results of other numerical models published in the literature. Parametric studies were carried out to assess the influence of material properties. It was found that D60, as a measure of gradation of aggregate base, and rainfall intensity and duration, are the most significant factors that influence the water migration within the pavement structures in the aftermath of simulated rainfall events. Subsequently, a computational process for numerical analysis that integrates the unsaturated hydraulic analysis with finite element structural performance was developed and verified against field FWD data. Using this computational tool, a comprehensive database of the results of numerical analyses with various permutations of input parameters was generated. This was achieved after several thousand runs by means of a high- performance computation facility. The generated input-output database was used to train several data-driven machine learning algorithms (e.g. Artificial Neural Network, Support Vector Machine, Random Forest Regressor, etc.). The advantage of these data-driven models is that they can rapidly estimate the structural response of a submerged pavement represented by a number of inputs without the need for performing cumbersome numerical analyses, which is critical for applications at network level. This dissertation provides insight into the performance of pavements subjected to major rainfall events and enables highway agencies to update their pavement management strategies accordingly