Decentralized modal analysis and system identification using embedded Markov parameter extraction within distributed wireless sensor networks

Abstract

As wireless monitoring systems continue to mature as a viable alternative to traditional wired data acquisition systems, scalable approaches to autonomously processing measurement data in-network are necessary. Embedded data processing has the benefit of improving system scalability, reducing the amount of wireless communications, and reducing overall power consumption. A system identification strategy based on Markov parameters is proposed for embedment within the decentralized computational framework of a wireless sensor network. Utilizing the computational resources of wireless sensors, individual sensor nodes perform local data processing to identify the Markov parameters of a structural system. Eventually, the global structural properties (e.g., mode shapes) are assembled by the wireless sensor network base station via an eigensystem realization algorithm executed using the limited number of Markov parameters transmitted by the wireless sensor nodes. The proposed strategy is evaluated using input-output and output-only data recorded during dynamic testing of a balcony structure.