Stochastic Sub-space Identification Methods for Bridges

Abstract

System identification of structures has drawn wide attention because of its potential applicability in many areas, such as structural vibration control [1] and vibration based health monitoring. Because structures such as bridges are most often excited with unmeasured ambient inputs, their identification requires a stochastic partial realization solution. In this paper, we explore this approach. We find that a vibrating structure can be modeled as a stochastic state space model whose modal (structural) parameters can be extracted from the identified system matrices. We study three experimental cases: a simulation study of a finite element beam model bridge, a simply supported steel beam, and data we collected at our test site at Walnut Creek Bridge on I-35 between Dallas, TX and Oklahoma City, OK. We find that our method can effectively identify the modal frequencies of the structures, and thus, the method can potentially provide useful data for the structural health monitoring of bridges.