New methods for inspection and evaluation of steel gusset plate connections

A new methodology was developed to collect accurate field measurements to support structural evaluations of steel truss bridge gusset plate connections. The method uses close-range photogrammetry to rectify field-collected digital images to produce scaled orthographic photographs (orthophotos) of bridge connections. Using the orthophotos, true-scale geometric measurements are made of the plate and fasteners. The geometric data can be compared with design and fabrication drawings and used to assess connection capacity. The methods and models have been deployed in the field and are finding broad acceptance. Introduction and Background Evaluation of gusset plate connections has become important for many transportation agencies in the US due to the recent collapse of the I-35W Bridge in Minnesota. According to the Federal Highway Administration, there are approximately 465 steel deck truss bridges within the National Bridge Inventory (NTSB Recommendations 2008). Even larger numbers of other types of steel truss bridges (approximately 12,600) exist in the inventory. Many of these bridges are undergoing additional scrutiny because the load paths are non-redundant, thus failure in a truss member or connection may cause the structure to collapse. Connection evaluations require complete and accurate as-built drawing sets and condition reports. Most connection evaluations use design drawings and traditional design methods to conduct ratings. Current methods to measure, collect, and archive field data are time consuming as illustrated in Fig. 1 and subject to errors at all stages. Additionally, field data will need to be archived to monitor and evaluate changes over the life of the structure. Sketches, notes, and qualitative photographic images may not be sufficient to provide definitive answers to time-dependent changes. Tools that can effectively capture field data to provide analysis inputs will hasten the complex and time consuming task of steel truss bridge connection evaluation. This paper reports on research that developed methods to create orthographic digital 1 Professor, School of Civil and Construction Engineering, Oregon State University, Corvallis, OR 2 Assistant Professor, Dept. of Civil Engineering, Istanbul Technical University, Istanbul, Turkey photographs (orthophotos) that enable metrification of steel truss bridge gusset plate connections. Data extraction from the images is ported directly to scripted Finite Element Analyses (FEA) to determine connection ratings. These combined techniques enable rapid and accurate quantitative field geometry acquisition and evaluation of connections. Integration of field data collection and analysis tasks further streamlines bridge management efforts. FIGURE 1 – EXAMPLE OF STATE-OF-THE-PRACTICE METHOD USED TO COLLECT FIELD DATA ON GUSSET PLATE CONNECTION GEOMETRY Visual inspection methods are now beginning to deploy supporting technologies that can improve and accelerate structural evaluations (McCrea et al. 2002). One type of technology is digital image processing, which has been utilized in various civil engineering fields. Several researchers implemented digital image technologies for assessment and inspection of steel, concrete and reinforced concrete structures. Although using digital image processing to detect a crack on a concrete surface is difficult due to voids, blemishes, shading, and shapes of cracks, it has attracted broad interest and been studied by several researches such as Ito et al. (2002), Dare et al. (2002), Hutchinson and Chen (2006), Fujita et al. (2006), Yamaguchi and Hashimoto (2006),Yamaguchi et al. (2008), and Yamaguchi and Hashimoto (2009). Lee and Chang (2005) used digital image processing for the assessment of rust defects on steel bridges. Liu et al. (2006) utilized image processing methods to detect rivets for aircraft lap joints. Simple and effective close-range photogrammetry techniques have been utilized in historical building documentation (Arias et al., 2007) and several researchers have used close-range photogrammetry for metrification such as Heuvel (1998), Criminisi et al. (2000), Tommaselli et al. (2005), Rodriguez et al. (2008). In this paper digital image processing is used to rectify digital photographs to produce scaled orthographic images (orthophotos) of steel truss bridge gusset plate connections so that physical dimensions can be extracted and used in connection evaluation. The present study differs from previous work because for evaluation of truss bridges, photographs cannot easily be taken from a mounted stationary position. In the field, photographs of bridge gusset plates will likely be taken from a snooper with both the snooper and bridge in motion from wind and traffic, or by climbing on the structure, and thus it will not be feasible to obtain stationary positions to correlate stereo or multi-station images. Image Rectification and Metrification with Flat-Field Lenses Photographs are taken of a real world image and placed on a two-dimensional image plane. When an image is captured with a camera and lens, it generally contains perspective distortion (parallel lines converging at a finite point), as well as other distortions due to the lens characteristics (such as barrel distortion or pin-cushion). To remove perspective, the image can be rectified using a mathematical transformation which maps elements in the real world image to those in the photographic image plane. To remove barrel distortion or pin-cushion, either flat field lenses must be deployed or lens correction parameters can be used to post-process images. For the present case, barrel distortion and pin-cushion are minimized by using lenses that minimize these distortions. When the salient features of the real world image correspond generally to a single plane, as is the present case for gusset plate connections, two-dimensional correspondences can be used to rectify the image, which simplifies the transformation. One of the most common techniques for image rectification is the direct linear transformation (DLT) algorithm. This transformation requires that certain geometrical characteristics be established between in the real world image and the image plane so that the image can be rectified. In the present case, point correspondences are used to map points between the real world image and the photographic image plane based on central projection shown in Fig. 2. Central projection maps the common points between planes and preserves lines in both planes.