Signal Separation of Simulated and Monitored Deflections Based on a Hybrid Bridge System Using the EEMD-GSA-LSSVM Approach

Abstract

Bridge health monitoring (BHM) technology serves as an effective tool in obtaining the state information of bridges. However, due to the limited budgets for maintenance of short- and medium-span bridges, BHM cannot be covered widely, although short- and medium-span bridges occupy a large proportion of global transportation networks. The original signals collected from BHM systems are composed of multiple components induced by vehicles, temperature, and environmental noise. It is thus important for bridge assessment to decompose the monitored data collected from the BHM system installed at short- and medium-span bridges. In this study, photogrammetric photoelectric transducer was used to monitor the deflections of a 170-m span hybrid bridge with spread steel box girders. A signal separation algorithm based on ensemble empirical mode decomposition (EEMD), least squares support vector machine (LSSVM), and gravitational search algorithm (GSA), named EEMD-GSA-LSSVM, is presented. Simulated mixed signals, which contain variety subsignals caused by vehicles, temperature, and noise, from the finite-element model were used to first examine the effectiveness of the mixed EEMD-GSA-LSSVM approach. It had been demonstrated that the mixed method successfully removed the environment-induced noise signals, and obtained the temperature-related deflection and dynamic deflection induced only by vehicle loads. The proposed method was finally applied to the actual single-day and multiday monitoring signals of the novel hybrid bridge in situ.