Research
1. Big Data Analytics in Support of Post-disaster Management
One of the critical tasks associated with the disaster management is to quantify the extent of damage and the possible impact on the community functionality. The time to recover some fraction of the pre-disaster infrastructure functionality, and the viable options to improve the performance are of prime importance after a natural hazard event. This research is focused on developing a robust framework using machine learning techniques to perform rapid risk assessments of building portfolios following a natural disaster (e.g. earthquakes). The resulting impact will be more effective rapid response tools to aid emergency responders in making effective and timely decisions. Currently, the method is being applied to buildings damaged during 2014 Napa Earthquake, and is proven to be effective.
2. Machine learning Techniques for Risk Assessment of Bridges
It is highly likely that various uncertainties such as geometric, material or component response parameters exist due to structure-to-structure variation in the generation of fragility curves, especially if the fragility curves are intended for the regional risk assessment of bridges. Also it is not possible/warranted to generate fragility curves for each bridge. This project is focused on developing a multi-parameter fragility framework using machine learning techniques. The proposed framework helps bridge owners (such as California Department of Transportation) to spend their resources judiciously (e.g. data collection, field investigations, censoring) on parameters that have a significant influence on bridge fragilities.
3. Grouping of Bridge Classes for Regional Risk Assessment
Fragility curves that can be applicable for vulnerability modeling of various configurations of structures, such as bridges, are required to facilitate regional seismic risk assessments. However, it is cumbersome and time-consuming to develop unique fragility curves for each structure across a regional portfolio. One strategy that has been adopted to address this challenge is to group bridges into classes with similar performance and fragilities. Traditionally, this grouping has been performed based on a relatively subjective identification of subclasses. However, such an identification leads to a number of bridge classes and unwarranted grouping. My current research explores various statistical grouping techniques such as Analysis of Variance (ANOVA), Analysis of Covariance (ANCOVA) to group bridge classes with statistically similar seismic demand.
4. Fragility curves: Traditional, Bayesian and Non-parametric approaches
The current study addresses the application of a traditional, Bayesian and non-parametric approaches towards the generation of parameterized (or bridge specific) fragility curves of bridges. The bridge-specific fragility curves can provide efficient and reliable fragility estimates of specific bridges with the same class in transportation networks without excessive computational efforts. In addition, one-dimensional fragility curves conditioned only on the intensity measure of the case study bridges can be developed by integrating the bridge-specific fragility functions over the domain of the uncertainty parameters to assess the vulnerability of bridge classes.
5. Force-deformation Characteristics of Masonry Spandrels
In unreinforced masonry walls, vertical piers are connected by horizontal elements called spandrels. These spandrel elements considerably affect the global wall behavior when subjected to seismic loading. This research explores the force-deformation characteristics of masonry spandrels with arches.