Automated assessment of tornado-induced building damage based on terrestrial laser scanning
Assessing the damage states of the built environment after a tornado is one method to better understand tornado induced loads and failure progress, as well as improve knowledge used for tornado mitigation, response, and recovery. Building damage states and geometry need to be measured and taken into consideration when estimating tornado wind pressures and speeds. This perishable damage data should be appropriately recorded, and investigations should be completed in a timely and unobtrusive manner in order to avoid interfering with clean up and recovery efforts that quickly change damage sites. Laser scanning or Light Detection And Ranging (LiDAR) provides 3D data that virtually captures damaged areas and allows for geometric queries and accurate measurements. Although robust, the 3D LiDAR data requires sophisticated data processing to extract meaningful information. The data processing becomes more challenging and time consuming when a large number of damaged buildings are investigated after a large-scale tornado. This research developed and tested a LiDAR data processing framework to automatically extract quantitative damage information needed for tornado wind speed estimation and structural damage analysis. The framework developed in this research includes methods for: 1) detecting damaged roof and wall surfaces in scans of damaged sites, 2) quantifying the percentages of roof/wall sheathing and covering losses, 3) estimating wind speeds at individual building scales, and 4) evaluating current tornado fragility models with actual damage information obtained by laser scanning. Performance of the developed methods was assessed with simulated data, laboratory scans, and actual data collected after large-scale tornadoes. A series of experiments in controlled conditions were conducted to determine the best algorithm settings and also objectively evaluate the performance of the proposed methods under varying environmental conditions. The proposed methods and their contributions to tornado wind speed estimation and structural fragility assessment were also tested with actual data collected after the Tuscaloosa, AL and Moore, OK tornadoes. The tests and case studies indicated that the detailed information extracted from LiDAR data could provide insight about tornado wind pressures and speeds with a resolution and accuracy not achievable with current visual inspections. This research determined the optimum data collection and processing settings that resulted in less than 10% error in calculating the percentages of roof and wall losses. The proposed method also estimates wind speeds at small-scale (individual buildings) as opposed to large-scale estimates often provided by field reconnaissance teams. Noting the fact that tornado wind speeds and pressures dramatically vary over short distances, the presented method provides engineers with a tool to improve the resolution and understanding of tornado effects, thereby improving building design and construction.