Aeroengine RUL Prediction with Time-Scale Invariant Latent Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional health indicator methods for aeroengine remaining useful life prediction neglect variability among different degradation processes, requiring additional normalization post-processing, which complicates the prediction process.
Innovation Solution
A neural ordinary differential equation (ODE) with symmetric regularization is used to model the degradation process in the latent variable space, incorporating time-scale transformations and symmetry regularization to constrain the neural network, enabling consistent latent variable processes without additional normalization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional health indicator methods are used to model degradation processes, then the model can be constructed with simple scalar indicators, but additional normalization post-processing is required to handle variability among different degradation processes
Solution Approach 1:
The patent applies preliminary action by incorporating normalization constraints directly into the model construction phase through symmetric regularization terms in the loss function. This prevents the need for post-processing normalization by addressing variability among different degradation processes upfront during training, thereby eliminating additional post-processing steps while maintaining model accuracy
2Adaptability or versatility
If traditional health indicator methods are used, then scalar indicators can represent system health, but variability among different degradation processes within a fleet is neglected requiring normalization
Solution Approach 1:
The patent applies parameter changes by transforming the scalar health indicator into a latent variable vector that captures multiple degradation features. The symmetric regularization term modifies the loss function parameters to enforce time-scale invariance, enabling the model to adapt to variability among different degradation processes within a fleet without requiring separate normalization procedures for each process
3Measurement precision
If neural ordinary differential equation is used to treat variability as time-scale transformations, then consistent latent variable structure can be learned, but symmetry regularization is required to enforce invariance
Solution Approach 1:
The patent applies feedback by incorporating a symmetric regularization term into the loss function that provides continuous feedback during training to enforce time-scale invariance. This regularization term monitors and corrects deviations from invariance constraints, ensuring the neural ODE learns a consistent latent variable structure that is invariant to time-scale transformations across different degradation processes
Data Source
AI summary
An aeroengine remaining useful life prediction method based on a neural ordinary differential equation under symmetric regularization includes the steps of: collecting monitoring parameters of a full-lifecycle of an aeroengine with sensors; establishing a first-order neural ordinary differential equation to perform continuous temporal modeling on a degradation process in a latent variable space, calculating a residual signal as a time-varying signal; establishing a Fourier neural operator to approximate a transfer function of a physical system, mapping an operating condition parameter and a latent variable to a sensor response parameter, so as to construct a loss function of the neural network; considering a time-scale transformation between different degradation processes, constructing a symmetry regularization term to constrain the invariance of the neural ordinary differential equation to the time-scale transformation, obtaining a latent variable process with a consistent structure.