In this work, the problem of fatigue damage prediction in the entire body of metallic structures through sparse output-only vibration measurements is investigated. The use of limited vibration measurements for output-only estimation of fatigue accumulation in structural systems was first proposed in [1]. Online strain estimates for multiple structural locations, derived from vibrational response measurements may be combined with S-N fatigue curves, damage accumulation models and rain-flow cycle counting procedures to predict accumulation of fatigue damage over the entire body of the metallic structure. In order to reliably predict fatigue damage, an accurate estimate of stress time histories is required, and the aforementioned necessitates high quality estimates of the states at corresponding degrees-of-freedom of the model. State estimation in presence of modeling uncertainties have been studied extensively, however, the input itself is typically assumed to be either known or broadband. In practice, the acquisition of precise load measurements is often impractical and sometimes impossible. Moreover, in operational cases the condition of stationarity is seldom satisfied. The latter substantiates the need to profit from the schemes developed for joint state and input estimation. Eftekhar Azam et al. [24] have recently proposed a novel dual Kalman filter (DKF) to accomplish the task of input-state estimation for linear time invariant systems via sparse acceleration measurements. In this article, the DKF and other state-of-the-art methods such as the augmented Kalman filter are employed for dynamic strain estimation enabling the development of fatigue damage accumulation maps. The effect of sensor configuration (number and location of sensors) on the accuracy of the fatigue estimates is also studied.
doi: 10.12783/SHM2015/296