Turboshaft Engine Degradation Prediction Under Varying Atmospheric Conditions

Solution Overview

Problem

Current data-based methods for aero-engine performance degradation prediction, particularly using neural networks, face challenges in stability and generalization, especially under varying atmospheric conditions, leading to unreliable multi-step predictions and increased errors over time.

Innovation Solution

The proposed method employs the Error Control Restricted Boltzmann Extreme Learning Machine (EC-RBELM) algorithm to establish prediction models for turboshaft engine performance parameters like gas turbine speed, power turbine inlet temperature, and specific fuel consumption, with offline learning and automatic updating of network topology parameters based on prediction errors, using Minimum Variance Weight (MVW) to improve model generalization across different atmospheric conditions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional neural network algorithms are used for engine performance degradation prediction, then the prediction can be performed without relying on accurate mathematical models, but the prediction stability is poor and results vary significantly across different tests due to random initialization of network parameters

Engineering Contradiction:

Improvemodel adaptability to different atmospheric conditionsVSAvoidprediction stability

Core Design Contradiction:

Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The patent applies preliminary action by using Restricted Boltzmann Machine (RBM) to pre-initialize the input layer parameters of the Extreme Learning Machine (ELM) before the main prediction task. This pre-initialization step prepares the network with better starting weights that are not purely random, thereby improving both stability and generalization performance across different atmospheric conditions.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent implements feedback through the Error Control mechanism that monitors prediction errors and uses them to adjust and optimize the network parameters. This feedback loop continuously improves the model's performance by learning from prediction discrepancies, enhancing both stability and adaptability to varying atmospheric conditions.

Inventive Principle:

Principle #23Feedback

2Productivity

If ELM algorithm with random input layer parameters is used, then the learning speed is fast and network structure is simple, but the generalization ability is poor under different atmospheric conditions and prediction errors accumulate in multi-step predictions

Engineering Contradiction:

Improvelearning speedVSAvoidprediction reliability in multi-step

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent applies preliminary action by using Restricted Boltzmann Machine (RBM) to pre-initialize the input layer parameters of the Extreme Learning Machine (ELM) before the main prediction task. This pre-initialization step prepares the network with better starting weights that are not purely random, thereby improving both stability and generalization performance across different atmospheric conditions.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent implements feedback through the Error Control mechanism that monitors prediction errors and uses them to adjust and optimize the network parameters. This feedback loop continuously improves the model's performance by learning from prediction discrepancies, enhancing both stability and adaptability to varying atmospheric conditions.

Inventive Principle:

Principle #23Feedback

3Stability of the object's composition

If RBM is used to initialize ELM input parameters to improve stability, then algorithm stability improves to some extent, but the generalization ability remains poor for data under other atmospheric conditions

Engineering Contradiction:

Improvealgorithm stabilityVSAvoidgeneralization ability under different atmospheric conditions

Core Design Contradiction:

Stability of the object's compositionVSAdaptability or versatility

Solution Approach 1:

The patent implements feedback through the Error Control mechanism that monitors prediction errors and uses them to adjust and optimize the network parameters. This feedback loop continuously improves the model's performance by learning from prediction discrepancies, enhancing both stability and adaptability to varying atmospheric conditions.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent applies dynamics by making the network parameters adaptive rather than fixed. The Error Control mechanism dynamically adjusts the RBM-initialized parameters based on prediction errors, allowing the model to adapt to different atmospheric conditions while maintaining the stability benefits of RBM initialization.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS20220300808A1Method for engine performance degradation prediction based on the EC-rbelm algorithm

Publication Date: 2022.09.22 NANJING UNIV OF AERONAUTICS & ASTRONAUTICS

AI summary

A method for engine performance degradation prediction based on the EC-RBELM algorithm, including establishing a prediction model cluster for three performance parameters, i.e., gas turbine speed Ng, power turbine inlet temperature T45 and specific fuel consumption SFC, in different atmospheric environments based on the EC-RBELM algorithm; learning EC-RBELM network topology parameters offline and automatically updating EC-RBELM network topology parameters based on prediction errors; and predicting the degradation of individual performance parameters of the turboshaft engine in different atmospheric environments according to the EC-RBELM algorithm model.