Leveraging data stream processing and weighted attack graph for real-time bridge structural monitoring and warning

Abstract

Regarded as one of the physical aspects under societal and civil development and evolution, engineering structure is required to support growth of the nation. It also impacts life quality and safety of the civilian. Despite of carrying dead load (its own weight) and live load during operation, structural members are also significantly affected by disaster and environment. Proper inspection and detection are thus crucial both during regular and unsafe events. An Enhanced Structural Health Monitoring System Using Stream Processing and Artificial Neural Network Techniques (SPANNeT) has been developed and is described in this paper. SPANNeT applies wireless sensor network, real-time data stream processing and artificial neural network based upon the measured bending strains. Major contributions include an effective, accurate and energy-aware data communication and damage detection of the already built engineering structure. Strain thresholds have been defined according to computer simulation results and the AASHTO (American Association of State Highway and Transportation Officials) LRFD (Load and Resistance Factor Design) Bridge Design specifications for launching several warning levels. SPANNeT has been tested and evaluated by means of computer-based simulation, test-bed and on-site levels. According to the measurements, the observed maximum values are 25 to 30 microstrains during normal operation. The given protocol provided at least 90% of data communication reliability. SPANNeT is capable of real-time data report, monitoring and warning efficiently conforming to the predefined thresholds which can be adjusted regarding user's requirements and structural engineering characteristics.