Eventos Anais de eventos
DINAME 2017
XVII International Symposium on Dynamic Problems of Mechanics
A Comparison Between Parametric and Non-Parametric Methods for Frequency Estimation of Blade Tip Timing Data
Submission Author:
Raphael Pereira Spada , SP
Co-Authors:
Raphael Pereira Spada, Rodrigo Nicoletti
Presenter: Raphael Pereira Spada
doi://10.26678/ABCM.DINAME2017.DIN17-0204
Abstract
Monitoring blade dynamics is an essential procedure to ensure proper performance of turbomachinery. For a long time strain gauges have been used with this goal, but the application of such sensors consists in a tedious, laborious, costly and time consuming task. Blade tip-timing is a technique developed in the 70’s to handle blade vibration monitoring in a non-intrusive way. Fundamentally, the methodology consists in placing sensors, usually optic, on a static casing around the turbine blades. Those sensors are responsible to give the passing time of the blades that are compared to a non-vibrating reference located at the shaft. By comparing the passing times of the vibrating blades with the non-vibrating reference, we can infer the blade vibrations. With the advances in software and hardware over the last decades, the research on this technique has gained momentum. Despite having several advantages over the strain gauge monitoring system, such as: easier installation; non-intrusive nature (doesn’t interfere with the blade dynamics) and is capable of measuring all the blades simultaneously, the technique has some major issues related to sampling frequency. This is due to the nature of the technique, because the sampling is directly linked to the rotating speed and number of sensors. Since the number of sensors is limited in practice, the resulting sampling frequency does not attend the Nyquist frequency criteria. As a result, aliasing is expected to occur in a spectral analysis of the signal. To attend to this matter, two types of approach have been developed along the years. The first, used to deal exclusively with synchronous vibrations, where the excitation is an integer multiple of the rotating speed, is based in parametric spectral analysis. The second is a more recent approach that is capable of dealing with both synchronous and asynchronous excitations and it is based on non-parametric spectral estimation. This paper presents a comparison between the performances of an autoregressive method (parametric) and a minimum variance spectral estimator with multi-sampling (non-parametric), applied to a simulated blade tip timing data, originated by a simple model of a bladed assembly. As a result, is shown that when dealing with synchronous vibrations the autoregressive method is very reliable, but being a parametric method, it requires some assumptions on the behavior of the structural model, including information on the type of excitation. On the other hand, the non-parametric method was able to perform blind analysis over the data but despite the developments made in the use of this technique, in the context of blade tip timing, some aliasing still occur even in the final spectrum obtained with the multi sampling method. Despite all, the non-parametric approach seems to be the one with most potential of application in real test cases, since no previous knowledge of the structure or assumptions were needed to be made.
Keywords
blade tip-timing, blade vibrations, spectral analysis
