Open Access Open Access  Restricted Access Subscription or Fee Access

Output-only Identification of Loading and Modal Parameters in Dynamic Systems with Non-Gaussian Inputs

T.J. ROGERS, K. WORDEN, R.S. MILLS, E.J. CROSS

Abstract


There exist engineering systems of interest which are not subjected to loads which can be modelled as a white Gaussian noise. A pertinent example of this is in the offshore energy industry where structures will be subjected to narrowband loading from the fluid structure interaction with the waves. This time-varying narrowband loading can make usual output-only identifications methods, e.g. stochastic subspace identification, difficult to implement effectively. The determination of modal parameters of a structure is a mainstay of modern SHM analyses, however, additional benefit can be found from recovery of the time history of the loading signal. This paper demonstrates the use of a Bayesian state-space formulation to address this problem. Rather than modelling the loads on the structure as a white-noise process, the loading is considered an additional latent state of the system. Additionally, that latent state is assumed to be a sample from a Gaussian process in time with a known stationary covariance function. In this way, a state-space representation of the Gaussian process governing the forcing is recovered. This approach enables the time history of loading on the structure to be recovered with quantification of uncertainty. The effectiveness of this method is demonstrated on a representative three-storey building structure.


DOI
10.12783/shm2019/32375

Full Text:

PDF