Search

Seismically vulnerable, especially collapse-prone, buildings often represent the greatest life-safety hazard worldwide. Identifying these buildings is the first step in seismic risk mitigation efforts for a given urban region’s resilience. This thesis aims to devise a workflow for the application of state-of-the-art deep neural networks (DNNs) for detecting and classifying seismically vulnerable buildings using three-dimensional point clouds. A number of prior studies have focused on using 2D image data in the field of structural health monitoring for damage recognition. The performance of those approaches for building classification at regional scales is highly dependent on well-controlled imagery data and may not be guaranteed when applied to real-world data. The present study, therefore, differs from prior studies in that it uses 3D point clouds, which implicitly contain depth information that can become a highly useful feature for training DNNs. Here, a specific DNN, namely, PointNet, is used for detecting soft-story buildings, which are ubiquitous in west-coast cities of the US. A workflow is devised for binary classification of point clouds into soft-story and non-soft-story points as well as the segmentation and association of classified point clouds with specific addresses. Over 10 billion point clouds obtained from the city of Santa Monica in California are manually labeled and split into training, validation, and testing sets, and the sensitive ranges of DNN hyperparameters are investigated to obtain the good performance.

The urban region's seismic resilience is being actively studied in recent years as a measure for risk mitigation, where the identification of seismically vulnerable buildings and the assessment of their performance play indispensable roles. However, it is a labor-intensive and computationally expensive task to evaluate tens of thousands of buildings in a region because the identification requires professional judgment at a site and the seismic assessment demands comprehensive modeling depending on structure-specific data. Nevertheless, it is feasible with the aid of advanced development of the Internet of Things (IoT) and computer technology. In this study, a data-driven framework including two pipelines that focus on soft-story buildings and non-ductile reinforced concrete frames is proposed.The first pipeline focuses on identifying soft-story buildings in the city of Santa Monica (California) through 3D point clouds and convolutional neural networks (CNNs). Although prior studies showed promising results in detecting soft-story buildings based on well-selected street-view images, false predictions are common when it is applied to real-world data. To address this issue, the pipeline implements point-cloud data where spatial information is available to segment building points and extract density features for training deep learning models and identifying soft-story buildings. The transfer learning (TL) technique is adopted to avoid overfitting in deep neural networks, and the parameters within the pipeline are investigated for optimal performance. The results illustrate the potential applicability of the pipeline for developing pre-and post-event countermeasures. The second pipeline focuses on another seismically vulnerable building, namely, the non-ductile reinforced concrete building (NDRCB). Prior studies indicated around 1,500 NDRCBs in Los Angeles that are urgently waiting for detailed assessment and mandatory retrofit or demolition if necessary. Because the fulfillment of these ordinances will last for decades, the potential risk of major losses will persist. To this end, an automatic method that harvests building information from IoT and imagery data generates archetypal models, conducts probabilistic seismic assessment, and estimates the losses for NDRCB frames is hence developed. The accuracy of the data harvesting module using deep CNNs is validated with the existing inventory data. The archetypal frames are developed based on the era-specific representative code and are validated through nonlinear static and nonlinear dynamic analyses of previously investigated NDRCBs. State-of-the-practice loss estimation methodologies including HAZUS and FEMA P-58 are adopted in the pipeline for constructing damage fragility functions and corresponding losses. The regional application focuses on intensity-based assessment for thousands of individual buildings instead of a scenario-based assessment. The outcomes of expected losses and repair/reconstruction time emphasize the vulnerability of NDRCBs in Los Angeles, and the presented pipeline is believed to bridge the gaps between property owners, engineers, and decision-makers.This research demonstrates how advanced data mining techniques and data-driven approaches can aid to solve civil engineering problems. While the framework currently focuses on soft-story and non-ductile frame buildings, it is expected to be extended in-depth and breadth in the future. That is, more detailed models and other seismically vulnerable infrastructures can be included.