PhD Final Defense – Casey Rodgers An Automatic and Quantitative Post-Earthquake Rapid Building Assessment Framework using Nonstructural Damage Meeting ID: 824 3534 3053 Password: 287276 Abstract Researchers have proposed the use of unmanned aerial vehicles (UAVs) for rapid and automated post-earthquake building safety assessment to address concerns with manual inspections.
These approaches typically assume that the UAV can obtain images of damage to the building’s structural components. However, in many cases, the structural system of a building and its corresponding structural damage are hidden from view underneath nonstructural components.
While these nonstructural components can be critical damage indicators that are often considered by inspectors during post-earthquake screening, most of the proposed automated inspection strategies do not consider damage to nonstructural components.
To this end, five research gaps have been identified: (1) The lack of an automatic post-earthquake rapid building assessment framework using only nonstructural damage; (2) The lack of a quantitative relationship between observed nonstructural damage and building safety condition; (3) The lack of a method that can automatically identify a wide range of nonstructural component types and materials; (4) The lack of a method that can automatically identify a wide range of nonstructural damage in specific damage states; (5) The lack of a platform to validate these methods.
To address these gaps, my overall research objective is to develop an automatic and quantitative post-earthquake rapid building assessment framework using nonstructural damage as a surrogate for building condition.
The research tasks are as follows: (1) Develop a strategy to quantitatively infer building safety condition from observed nonstructural damage; (2) Automatically identify and classify nonstructural components according to FEMA P-58 by creating new methodologies for instance segmentation of nonstructural components; (3) Automatically identify nonstructural damage according to FEMA P-58 and Hazus by creating new methodologies for instance segmentation of nonstructural damage; (4) Create a validation platform for generation of architecturally accurate and realistic synthetic buildings and nonstructural damage data; (5) Integrate tasks 1-3 into one end-to-end framework and validate the framework.
This research will develop and utilize many techniques in statistics, computer vision, deep learning, computer graphics, and civil engineering. The completion of these five extensive tasks will lead to an automatic and quantitative end-to-end framework for post-earthquake rapid building assessment using only nonstructural damage, including a quantitative relationship between observed nonstructural damage and building safety condition, instance segmentation and classification of nonstructural components and damage, and procedural generation of architecturally and structurally accurate and realistic nonstructural components and damage.
This framework will allow inspectors to automatically perform quantitative post-earthquake rapid building assessments, even if the structural system is hidden from view, reducing the time needed to inspect all buildings after an earthquake and allowing people and communities to heal faster after a tragedy.