Improvement of the accuracy of nonlinear rotordynamic models by means of the use of non-white fluid-induced signal noise

Abstract

When performing model simulations in the field of rotordynamics the classical approach for production and subsequent analysis of noisy signals is the addition of white noise to the numerical results. Nevertheless, the noise spectrum observed in monitoring signals from operating rotating machinery rarely resembles white noise; this can be troublesome for noise-reduction filters, usually tuned to deal with white noise. In these machines, signal noise is originated mainly from the interaction of mechanical components with the working fluid flow. Considering this, this manuscript presents a method to simulate the mechanical response of a selected rotor system model to fluid-induced pressure fluctuations, thus increasing modeling accuracy in systems under turbulent-fluid effects. It is supposed that wall pressure fluctuations behave according to the empirical models of Corcos and Goody. A sample of fluid-induced signal noise generated with this method is used to evaluate the robustness of signal processing methods for the detection of single-point rub in aeroderivative gas turbines, while proving how signal features may be affected differently by diverse types of noise. Finally, we present an experimental validation of the methodology on a setup that was adapted to host an axial flow of compressed air throughout the rotor-casing assembly.