Dynamic response predictions for a mistuned industrial turbomachinery rotor using reduced-order modeling

R. Bladh, C. Pierre, M. P. Castanier, M. J. Kruse

Research output: Contribution to journalArticlepeer-review

150 Scopus citations

Abstract

This paper explores the effects of random blade mistuning on the dynamics of an advanced industrial compressor rotor, using a component-mode-based reduced-order model formulation for tuned and mistuned bladed disks. The technique uses modal data obtained from finite element models to create computationally inexpensive models of mistuned bladed disks in a systematic manner. Both free and forced responses of the rotor are considered, and the obtained results are compared with "benchmark" finite element solutions. A brief statistical study is presented, in which Weibull distributions are shown to yield reliable estimates of forced response statistics. Moreover, a simple method is presented for computing natural frequencies of noninteger harmonics, using conventional cyclic symmetry finite element analysis. This procedure enables quantification of frequency veering data relevant to the assessment of mistuning sensitivity (e.g., veering curvatures), and it may provide a tool for quantifying structural interblade coupling in finite element rotor models of arbitrary complexity and size. The mistuned forced response amplitudes and stresses are found to vary considerably with mistuning strength and the degree of structural coupling between the blades. In general, this work demonstrates how reduced order modeling and Weibull estimates of the forced response statistics combine to facilitate thorough investigations of the mistuning sensitivity of industrial turbomachinery rotors.

Original languageEnglish
Pages (from-to)311-324
Number of pages14
JournalJournal of Engineering for Gas Turbines and Power
Volume124
Issue number2
DOIs
StatePublished - Apr 2002

Fingerprint

Dive into the research topics of 'Dynamic response predictions for a mistuned industrial turbomachinery rotor using reduced-order modeling'. Together they form a unique fingerprint.

Cite this