Comparative Study of Parameterizations for Damage Localization with Finite Element Model Updating

With this work, we present a comparative study of parameterization methods in finite element (FE) model updating with the goal to localize damage in a wind turbine rotor blade. The choice of design variables greatly impacts the quality of the model updating procedure. A common approach is to determine geometric regions where the probability of an emerging defect is known to be high, based on experience or prior knowledge. Then, mechanical properties of these susceptible regions are directly fitted to measured behavior in order to find the position where damage has occurred. A large number of such regions can result in an objective value space with many local minima, making numerical optimization unfeasible. To alleviate this problem, we introduce a stiffness (i.e. damage) distribution function that is described by only few parameters. By employing a cumulative distribution function, the proposed parameterization is independent of the FE mesh resolution as well as of prior assumptions about the defect location. We compare the proposed parameterization to the commonly used method, that directly uses mechanical properties of geometric regions as design variables. We employ the two parameterizations in FE model updating of a typical wind turbine rotor blade, where we introduce a fictitious defect to simulate the target state.