Automated Characterization of Bridge Deck Distress Using Pattern Recognition Analysis of Ground Penetrating Radar Data
Scott, Michael Leon
MetadataShow full item record
Many problems are involved with intspecting and evaluating the condition of bridges in the United States. Concrete bridge deck inspection and evaluation presents one of the largest problems. The deterioration of these concrete decks progresses more rapidly than any other bridge component, which leads to early concrete deck replacements that must be done before the bridge superstructure needs to be replaced. The primary cause of deterioration in these concrete bridge decks is corrosion-induced concrete cracking, which frequently results in delaminations. Delamination distress increases the life cycle cost of maintaining a concrete bridge deck, particularly when it is not detected early on. Early detection of delamination distress can facilitate economical repair and rehabilitation work, but bridge engineers must recommend deck replacement if repairs are delayed too long or inspection tools cannot detect delaminations early enough. The Federal Highway Administration has responded to the need for a better bridge deck inspection tool by contracting Lawrence Livermore National Laboratory to develop two new prototype ground penetrating radar systems. These two systems generate three-dimensional data that provide a representation of features that lie below the bridge deck surface. Both of these systems produce large amounts of data for an individual bridge deck, which makes automated data processing very desirable. The primary goal of the automated processing is to characterize bridge deck distress represented in the data. This study presents data collected from sample bridge deck sections using one of the prototype systems. It also describes the development and implementation of appropriate methods for automating data processing. The automated data processing is accomplished using image processing and pattern recognition algorithms developed in the study.
- Doctoral Dissertations