Maxime Koebelé-Cousquer

Dipl.-Ing. Maxime Koebele

  • Postanschrift:
    Karlsruher Institut für Technologie
    Institut für Technische Mechanik
    Teilinstitut Dynamik/Mechatronik
    Postfach 6980
    76049 Karlsruhe

    Haus- und Lieferanschrift:
    KIT-Campus Süd
    Institut für Technische Mechanik
    Teilinstitut Dynamik/Mechatronik
    Geb. 10.23, 2.OG
    Kaiserstraße 10
    76131 Karlsruhe

     

     

Stochastic Analysis of Geometric Mistuning in Radial Compressors

radial compressor
Histogram of the maximum blade response amplitude obtained by Monte Carlo simulation
Sector of an academic bladed disk with random geometry modifications

Radial compressor in turbochargers is often considered in theory as periodic system, but in fact it features inevitable small imperfections caused by material defects and manufacturing tolerances which break the cyclic periodicity. This is called mistuning. The loss of periodicity changes drastically the dynamic behavior of the compressor. Typically the forced response level of the mistuned bladed disk is larger than the tuned design. Because of the random nature of mistuning, the determination of the largest resonant response at any frequency has to be considered as a stochastic problem.

Mistuning receives significant attention from the research community since the late 1960s. Models using coupled lumped mass oscillators have allowed the fundamental phenomena of mistuning to be understood. More recently finite element model are used to explore with a better precision the behavior of mistuned compressor. To minimize the computational costs the finite element model has to be reduced first before performing a Monte Carlo Analysis. In the last decade several model reduction methods were developed. The way in which mistuning is implemented depends on the used reduction method and the vast part of them accounts only for a frequency mistuning model or a proportional mistuning model, in which the ideal tuned configuration is not modified.

Today a key topic is to combine this reduction methods with a more realistic introduction of mistuning, such as geometric mistuning. In other words, it will be tried to take directly random geometry modifications into consideration in a stochastic problem.

Contact: Prof. C. Proppe, M. Koebele

Publikation