Dynamic response of an industrial turbomachinery rotor

This paper investigates the effects of random blade mistiming on the dynamics of an advanced gas turbine rotor. Both free and forced responses of the rotor are examined using the finite element method, and a computationally inexpensive reduced-order modeling technique based on component mode synthesis. The spatially extended free modes of vibration of the tuned rotor are found to undergo severe localization upon the introduction of blade mistuning. In turn, this results in dramatic displacement and stress amplitude increases in the forced response of individual blades. The mistuned forced response amplitude is found to vary considerably with mistuning strength and the degree of aerodynamic and disk structural coupling between the blades. The paper concludes with a statistical study in which Weibull distributions are used to calculate approximate forced response statistics. predicted by a tuned analysis.' Mistuning effects must be included in the analysis if accurate predictions of vibratory response amplitudes are to be made. However, analyzing a finite element model of a full blade assembly, such as that shown in Figure 2, is an enormously costly, if not impossible, computational task. The purpose of this paper is to demonstrate, through the case study of an industrial rotor, that mistuned response amplitudes can be accurately and efficiently predicted, hi a statistical fashion, by using a novel reduced-order modeling technique. This technique, recently developed by Ottarsson et a/. and Kruse and Pierre, produces reduced-order models (ROM) of turbomachinery rotors directly from their finite element models. The procedure involves a component mode analysis of the rotor, with a truncated number of modal amplitudes describing the response of the assembly. The key idea introduced by Ottarsson et a/. is that the motion of an individual blade consists of both cantilever blade elastic motion and diskinduced static motion. The principal advantage of the reduced-order modeling technique is the considerable computational savings associated with solving for the dynamic response of an entire mistuned rotor