Vibration detection in turbomachinery using non-contacting sensors

Abstract

Recent developments have seen the introduction of multiple Eddy Current Sensors (ECS) into turbomachinery. These sensors employ an active magnetic field to monitor each blade as it passes the sensor. They generate an electrical signal proportional to the distance of a blade from the sensor. Existing algorithms extract two pieces of information from the ECS signature, signal magnitude and signal zero crossing time. The signal magnitude is used to find tip clearance, and the zero crossing time is used to estimate vibrational parameters over the course of multiple revolutions. These techniques fail to exploit the majority of the information contained in the ECS signal. In this research, a novel residue characterization algorithm was developed that processes the full ECS pulse to produce a residue. The residue is a speed independent representation of differences between a baseline ECS pulse and an experimental ECS pulse. A mathematical model of the relationship between blade displacement and residue was developed. Empirical data collected with the MIT spin pit were used to verify convergence of the model with the residue characterization algorithm. This strongly suggests blade vibration can be deduced using the residue characterization method with a single ECS sensor.