Abstract:

To ensure the safety and longevity of infrastructure, Visual Inspection (VI), Nondestructive Evaluation (NDE), and Structural Health Monitoring (SHM) technologies are used to assess performance and condition of our nation’s highway infrastructure, and quantify the deterioration. Advances in sensor technologies, instrumentation and data systems have allowed us to collect data safer, faster and more reliably. While such data has become easier and cheaper to collect with advanced sensing technology and automated data acquisition, the ability to efficiently process data (without extensive expertise), extract actionable information, fuse disparate data sources, and develop visualization tools that permit intuitive interpretation remain elusive due to the complexity, uncertainty, and heterogeneity of infrastructure assessment data. The Federal Highway Administration (FHWA) initiated a study to develop data fusion strategies and algorithms that enhance the ability of the data visualization scheme to convey information related to the condition, performance and safety of bridges. In this study, visualization and fusion schemes for NDE, SHM and VI data are investigated, and a holistic visualization scheme will be presented that (1) conveys meaningful and actionable information about the infrastructure that would trigger an intervention, and (2) is understandable by the relevant individuals (such as state DOT engineers and consultants without extensive background in NDE and SHM). This paper documents the state of data fusion being used in other fields that may be adopted and refined for highway infrastructure, as well as benefits and limitations of different fusion and visualization techniques with respect to the type of data collected and type of actionable information provided to the owners for bridge inspection and management. The paper also documents States’ perspective and pressing needs on data fusion and visualization. The findings suggest that the key challenges for infrastructure assessment are the small size of existing data sets combined with the complex measurement and uncertainty of such data. In tandem, these challenges prevent the adaptation of many conventional data assimilation approaches, and suggest a need for nonparametric and machine learning-driven fusion and visualization. Additionally, this study indicates that biomedical analysis holds many similarities with infrastructure assessment, and consequently that visualization and data analysis techniques developed in that domain hold potential for adaptation.