The location method of blade vibration events based on the tip-timing signal

Abstract

Tip-timing technology has been widely used to monitor blade vibration of the aeroengine. In the off-line analysis of tip-timing signals, it is key and a prerequisite in blade fault diagnosis to locate the vibration event accurately. It is the most common method used to locate abnormal vibration based on the correlation of tip-timing data in adjacent revolutions. However, the data in the adjacent revolutions only include little vibration information, which results in the location performance being susceptible to noise. This paper located the vibration event using the area of the ellipse calculated based on the two-parameter plot. The relationship between the area of the fitted ellipse and the blade vibration parameters was derived for the first time in this paper. The feasibility of the method was verified using the tip-timing data of the low-pressure fan of an aeroengine. The results showed that the method can locate both synchronous resonance and rotating stall accurately. Its performance in anti-noise interference was far superior to the correlation coefficient methods due to enough information provided by multi-revolutions rather than only two revolutions. The work in this paper is of great significance for the realization of automatic processing of tip-timing signals.

Abnormal time sequence identification algorithm for power measurement devices based on variational autoencoder and support vector machine

“…Among them, time series anomaly detection methods based on prediction model the historical normal behavior of multivariate time series, construct regression models, predict the corresponding values of the next time, define anomaly scores based on the difference between the predicted values and the true values to determine whether the data is abnormal. Although these methods can learn the normal patterns of historical time series and predict the future trend of time series data changes [6] , they are easily affected by noise, Simultaneously, there is a large amount of computation; The core idea of reconstruction based temporal anomaly detection methods is to learn the compressed representation of normal temporal data patterns, and then use it to reconstruct the original input. The anomaly score is defined by the difference between the reconstructed value and the input value to determine whether the data is abnormal.…”