Measurement Point Selection and Damping Identification of Blisks.

Abstract

Capturing the motion of an integrally bladed disk or blisk can be very difficult and typically involves finite element models with a large number of degrees of freedom (DOFs). These models employ parameters which are often not well known, for example the damping. Thus, identification techniques are needed to determine the actual damping.

Due to wear or manufacturing, nominally cyclically symmetric blisks have slight variations in the mass or stiffness of their components known as mistuning. As a result, the cyclic symmetry is destroyed and vibration energy can be localized around certain regions of the system leading to a larger than expected forced response as compared to the response of the analog cyclically symmetric (or tuned) structure. As a result, the mistuned structure is more susceptible to high cycle fatigue and earlier failure than the tuned structure. Damping plays an important role in investigating the effects of localization, because damping affects the forced response of a mistuned system (in particular, it affects the maximum response amplitude).

Current damping identification methods often have diffculty for regions of high modal density. Also, they typically require knowledge of complex eigenvalues and eigenvectors, the actual applied forcing, or energy measurements. Current methods assume that accurate measurement data has been measured, but they do not provide information on how this assumption is realized. This work introduces a measurement point selection method which results in an accurate system identification with minimal experimental and computational cost.

In addition, this work proposes new damping identification methods for structural, viscous modal, and component damping models. Addressing existing challenges of current damping identification methods, the proposed methods apply to systems with low or high modal density (such as mistuned blisks), only require knowledge of the forced response, the relative forcing, the mistuning, and a finite element model. Also, they can use either full order or reduced-order models. In addition, these methods are shown to be accurate in the presence of measurement noise and to capture the system dynamics for a validation blisk.