Method and device for predicting residual service life of aero-engine through time-frequency domain analysis

Abstract

The invention provides a time-frequency domain analysis aero-engine remaining service life prediction model method and device, and relates to the technical field of aero-engine service life prediction and health management. Comprising the following steps: preprocessing aero-engine sensor data, carrying out continuous wavelet transform to obtain time-frequency domain characteristics, and constructing network input from three dimensions of time, a sensor and the time-frequency domain characteristics; and establishing and training a 3DConLSTM neural network to obtain an output result, namely prediction of the remaining service life. According to the method, continuous wavelet transform is introduced, and historical data of the aero-engine are analyzed in a time domain and a time-frequency domain respectively. The characteristics of the LSTM network are adopted, three gating units control inflow of information, and the long-term degradation trend of time sequence data is reserved; according to the method, dot product operation in a traditional LSTM network is changed into convolution operation, features among three-dimensional inputs are effectively extracted, time sequence and spatiality of data are considered, the residual service life prediction result is more accurate, and the residual service life prediction precision is improved.

Description

technical field [0001] The invention relates to the technical field of aero-engine life prediction and health management, in particular to a time-frequency domain analysis method and device for predicting the remaining service life of an aero-engine. Background technique [0002] RUL (Remaining Useful Life) prediction can predict the time from the operating state of the equipment at a specific moment to the fault state. Existing RUL methods mainly include three categories: physical model-based methods, data-driven methods and hybrid methods. First, physical model-based methods develop physical failure models by studying the internal interpretation mechanisms of equipment. These methods are developed on the basis of physical system knowledge, and the accuracy is limited when the system is complex. Second, data-based methods can be based on historical sensor The data learns potential degradation trends. Third, hybrid methods try to integrate the above two methods, but sometim...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the problem that CNN ignores the timing of data in the prior art, and the RNN network does not analyze and consider the correlation between multiple input features, the present invention proposes a time-frequency domain analysis method for predicting the remaining service life of an aero-engine and device

Images