Reduced-Order Model Construction Procedure for Robust Mistuning Identification of Blisks

The reduced-order modeling of integrally bladed disks for predicting the mistuned vibration response has been well studied and understood. For solving a direct vibration problem, adding modes to the modeling basis improves the accuracy of the reduced-order model with respect to the parent finite element model. In contrast, when solving an inverse problem for system identification, adding modes to the reduced-order model while using the same measurements may actually reduce its accuracy. This is especially true for solving inverse problems related to the identification of blade mistuning parameters, because the characteristics of the selected system modes for the reduced-order model may not match the assumptions used in the mistuning modeling approach. In this work, a procedure is introduced for constructing a reduced-order model referred to as the inverse reduced-order model that is well suited for solving the mistuning identification inverse problem. First, a quantitative metric is defined to characterize and rank the tuned-system modes with respect to their suitability for constructing inverse reduced-order models. Then, the direct problem is solved using a larger direct reduced-order model with prescribed mistuning to interrogate and validate the performance of various inverse reduced-order models as modes are added. This enables the automated construction of suitable inverse reduced-order models and improves the overall accuracy and robustness of mistuning identification.