RUL prediction of electronic controller based on multiscale characteristic analysis

The reliability of an electronic controller, which is usually determined by analysing its performance degradation under the working conditions and environmental stresses, has a significant impact on aircraft engine safety. In this paper, a hybrid degradation model, which combines multiscale characteristic analysis (MCA) with modified Gaussian process regression (GPR), is proposed to predict the remaining useful life (RUL) of a controller under various working conditions. Ensemble empirical mode decomposition (EEMD) is utilized to decompose the original data into a number of independent intrinsic mode functions (IMFs) consisting of both degradation and fluctuation information. Characteristic analysis, including series extraction and importance measurement, is conducted to identify the main characteristics hidden in the target IMF. To describe the degradation path under various working conditions, an equivalent function describing the relationship between the degradation rate and stress levels is constructed. It is applied to modify the mean function of the GPR model to describe the relationship between the input of the time series and the degradation trend extracted via EEMD. Meanwhile, in the modified GPR model, combinations of three kinds of covariance function are used to capture the scale characteristics of periodicity and mutability. Then, these individual GPRs are aggregated into the final degradation model. Based on that, the prognostic probability distribution of a controller's RUL can be calculated numerically via a Monte Carlo (MC) simulation. Finally, the effectiveness and accuracy of the proposed method are verified.