Session Abstracts:
Session 7A
Postdoctoral Scholar
University of California, Berkeley
Presentation Title: NHERI SimCenter Workflows for Automated Extraction of Inventory and Damage Data from NHERI RAPID Reconnaissance Data
Abstract: The last few decades have seen major strides in the modeling and simulation capabilities required for natural hazards engineering applications. Despite these significant advancements, the building and infrastructure inventory data needed to effectively use these new capabilities toward granular regional-level studies remained largely unavailable, which undermined the applicability of these new methods to large-scale problems. The recent work utilizing deep learning (DL) and computer vision (CV) models on publically available remote sensing data to extract inventory data shows great promise in tackling the inventory data gap. This presentation will focus on one of the few successful and scalable implementations of DL/CV-based inventory extraction, i.e., NHERI SimCenter's BRAILS pipeline. Specifically, the presentation will discuss the recent findings showing BRAILS' ability to better extract inventory information using high-fidelity reconnaissance data collected by NHERI RAPID and the current methods built into BRAILS to use RAPID data in establishing granular building inventory data. Another focal point of this presentation will be to showcase potential uses of RAPID data for automated damage detection using BRAILS' DL/CV models. A brief discussion of the future BRAILS enhancements for utilizing RAPID data for further meeting natural hazards engineering needs will also be presented.
Associate Professor
Oregon State University
Presentation Title: Estimation of the behavior of a corroded steel industrial building using lidar generated section properties
Abstract: Corrosion is one of the major causes of structural deterioration for steel structures. For steel structures, corrosion deterioration can lead to material degradation, resulting in a reduction of material strength and stiffness properties and ultimate failure of the structure. Current assessment of corroded steel is highly variable and often leads to inaccurate evaluation of the structure’s condition. To overcome these limitations, a quantitative assessment methodology was developed using remote scanning techniques coupled with finite element method (FEM) modelling. In this study, the methodology was applied to a two-story corroded steel-framed structure located on a historical industrial site. The geometric information and cross-sectional properties of the structure were obtained from lidar point cloud data, which were then used for FEM analysis to evaluate the residual seismic behavior of the building.
Professor
University of Washington
Presentation Title: Characterizing Damage to a Full-Scale Reinforced Concrete Building Tested using lidar
Abstract: The collection of time-sensitive data and the rapid and reliable assessment of structural safety are paramount in the aftermath of natural disasters. This presentation discusses the calculation of residual inter-story drifts and identification and measurement concrete cracking and spalling using lidar, in the context of shake table testing of a full-scale reinforced concrete building. Several examples are used to illustrate that point cloud data can indeed be processed to obtain measurements of global residual drifts or of the extent of visible damage (e.g., crack width, spalling) in specific structural components. The results presented indicate that collecting post-disaster data using lidar scanners has the potential to improve how structures are assessed in the aftermath of seismic events.
PhD Student
Oregon State University
Presentation Title: Virtual Damage Assessment of Buildings Impacted by Hurricane Ian (2022) in Fort Myers Beach, FL
Co-Authors: Mehrshad Amini; Daniel T. Cox; Andre R. Barbosa
Abstract: In September of 2022, the town of Fort Myers Beach (FMB), located 14 miles southwest of Cape Coral on Estero Island, was directly impacted by Hurricane Ian. The storm surge covered the low-lying barrier island completely, leading to high levels of building damage. Multiple academic and professional groups collected hurricane damage imagery at FMB, including the Natural Hazards Engineering Research Infrastructure (NHERI) Structural Engineering Extreme Event Reconnaissance (StEER) team, the National Oceanic and Atmospheric Administration (NOAA), the Federal Emergency Management Agency Mitigation Assessment Team (FEMA-MAT), the American Society of Civil Engineers (ASCE), Oregon State University (OSU), the Florida Department of Environmental Protection (FDEP), and others. These data were used to conduct a robust hindcast analysis in the form of a virtual damage assessment (VDA). This study involved the development of a VDA framework used to train 18 undergraduate students to conduct a component-based damage assessment of 3,408 buildings on FMB, with overall damage states (DS) ranging from no damage (DS0) to complete (DS6). The study concludes that the VDA methodology is effective and accurate to determine the DS of impacted buildings. The results were validated through a cross-comparison of student and expert assessments that, along with feedback sessions from both groups, confirmed the expected level of uncertainty in the DS observations. With over 3,000 buildings evaluated, this study is a first step in addressing the data gap that currently exists for the validation of damage models and of artificial intelligence-based damage assessment tools.