Fault prediction method for aero-engine

Abstract

The invention discloses a fault prediction method for an aero-engine, and the method comprises the steps: taking the data of a complete life cycle of an engine as much as possible, carrying out the anomaly detection of the data, drawing the feature recession trajectories of all life cycles, carrying out the fitting of the recession trajectories of all complete life cycles, and obtaining a most universal recession curve; then fitting the recession curve through an improved time sequence analysis algorithm to obtain an algorithm model. Compared with a traditional time sequence analysis algorithm, the improved time sequence algorithm can predict future time which can be more than 10 times that of the traditional time sequence algorithm under the condition that the prediction accuracy is the same.

Description

technical field [0001] The patent of the present invention relates to the aerospace field, relates to the construction technology of an aero-engine fault prediction system, and specifically relates to an aero-engine fault prediction method. Background technique [0002] Compared with ordinary data, aeroengines have the characteristics of large data volume, fast data generation (up to 5HZ per second), large data fluctuations, and complex working conditions. Conventional fbprophet, ARIMA, and LSTM timing algorithm processing ideas are difficult to directly apply to aero-engine data. Although the characteristics of aero-engines change regularly in multiple cycles, the flight conditions at different times are all operated by the pilot. Different flight conditions will lead to different changes in the characteristics. Therefore, even if the timing algorithm can predict The value of the feature, but the predicted value obtained by this prediction method does not have any referenc...

AI Technical Summary

Problems solved by technology

Conventional fbprophet, ARIMA, and LSTM timing algorithm processing ideas are difficult to directly apply to aero-engine data

Although the characteristics of the aero-engine change regularly in multiple cycles, the flight conditions at different times are all operated by the pilot. Different flight conditions will lead to different changes in the characteristics. Therefore, even if the timing algorithm can predict The value of the feature, but the predicted value obtained by this prediction method does not have any reference point value in the actual production environment

Images