Session: 05-05: Vibration Monitoring Analysis

Paper Number: 154063

Compressor Blade Vibration Measurement Using Radio Frequency Amplitude and Frequency Modulation 

Repeated cycles of forced blade vibrations cause high cycle fatigue, a leading cause of component failure in modern gas turbine engines. Current engine maintenance is schedule-based, where engines are overhauled based on the number of flight hours. This may cause some engines to be overhauled while they are still in good health, whereas other engines may fail before the next inspection. The alternative,  condition-based maintenance, is generally accepted as an effective way to reduce maintenance costs and catastrophic failures. However, condition-based maintenance requires in-flight engine information. There is currently no technology available to collect blade vibration data in flight. 

Radio Frequency (RF) sensors have the potential to provide in-flight engine health information. The proposed RF sensor system consists of waveguide antennas which are non-contact, low-profile, low-power and well suited for implementation in a gas turbine engine environment. A series of experiments were conducted in a single-stage compressor to study the RF sensor response to blade vibrations. Reciprocal dual-polarized 18 GHz waveguide antennas were placed over the rotor, while synchronous and asynchronous vibrations were induced. Data was simultaneously collected with the RF sensors and with a Nonintrusive Stress Measurement System (NSMS). 

A mathematical model based on amplitude and frequency modulation is proposed. The RF signal is amplitude modulated by the blade passage and blade vibrations, and frequency modulated due to Doppler shift from blade passage. The model showed qualitative agreement with the RF data in the frequency domain. Amplitude and frequency modulation causes asymmetrical sidebands around the RF frequencies. The sideband amplitudes showed a correlation with the blade vibration ampltiudes. The sideband amplitudes were used as the inputs into a multi-layer neural network. The outputs of the neural network were the blade vibration amplitudes by nodal diameter. The neural network was tested with RF data for an acceleration through a synchronous crossing and tested with data for a deceleration through the synchronous crossing. Deflection predictions from the neural network agreed with the NSMS deflection measurements to within 130 microns for a 1100 micron peak-to-peak amplitude deflection. 

Presenting Author: Yuko Inoue University of Notre Dame

Presenting Author Biography: Yuko Inoue is a PhD candidate at the University of Notre Dame. Her research interests are turbomachinery instrumentation, aeromechanics and artificial neural networks.

Authors:

Yuko Inoue University of Notre Dame
Nathaniel Griggs University of Notre Dame
Thomas Pratt University of Notre Dame
Scott Morris University of Notre Dame