Time-Domain Parameter Estimation Algorithm for Structures. II: Numerical Simulation Studies

Abstract

Monte Carlo simulation is used to study the statistical behavior of the time-domain parameter estimator presented in the companion paper, using a bridge truss as a model problem. The effects of the estimation time step, the initial parameter values, and the quality and quantity of information on the outcome of parameter estimation are investigated. The paper demonstrates that the time step must be sufficiently small to control the errors inherent in the numerical integration scheme, but should be large enough to provide new information for the estimator. As expected for this type of problem, the estimates are biased in the presence of noise in the measurements. However, for practical amplitudes of noise, the estimated parameters are sufficiently reliable. The estimator has the desirable characteristic that, when the noise amplitude is sufficiently small, the accuracy of the estimates can be increased by presenting more information to the algorithm.