Gas Turbine Inlet Condition Estimation With Neural Model Compensation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional aircraft gas turbine engine control systems face inaccuracies due to inlet condition sensor faults, leading to incorrect engine control and potential failures, especially during transient and specific operating regimes, which existing models struggle to accurately address.

Innovation Solution

A system utilizing an aero-thermodynamic model augmented with machine learning, specifically neural networks, to estimate and correct engine inlet conditions, incorporating fault detection and accommodation to improve model accuracy and adapt control laws for various operating regimes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If conventional piece-wise linear state variable models are used for real-time engine control, then computational burden is reduced and real-time deployment is enabled, but model accuracy deteriorates during transient and specific operating regimes

Engineering Contradiction:

Improvereal-time control capabilityVSAvoidmodel accuracy during transients

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies dynamics by making the model structure adaptable and changeable based on operating conditions. The neural network compensation dynamically adjusts model predictions during different operating regimes, allowing the system to switch between simplified real-time operation and enhanced accuracy modes as needed

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent creates a composite modeling approach by combining the piece-wise linear state variable model with neural network compensation. This hybrid structure integrates the computational efficiency of simple models with the accuracy enhancement of machine learning, resolving the contradiction between real-time capability and transient accuracy

Inventive Principle:

Principle #40Composite materials

2Measurement precision

If adaptable modeling techniques with Kalman filter observers are used to improve model accuracy, then steady state matching is improved, but model deficiency during transient operations persists and tuner levels become unreasonable

Engineering Contradiction:

Improvesteady state model accuracyVSAvoidmodel reliability during transients

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent introduces neural network compensation as an intermediary layer that mediates between the base model predictions and actual sensor measurements. This intermediary specifically addresses transient operations where the base model fails, providing correction without requiring unreasonable tuner adjustments

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the approach from adjusting physical tuner parameters (efficiencies, flow parameters) to using neural network learned parameters. This allows the model to adapt to transient conditions through data-driven parameter adjustments rather than physically unreasonable tuner levels

Inventive Principle:

Principle #35Parameter changes

3Device complexity

If inlet condition sensor faults are not detected and accommodated, then system complexity remains low, but engine control accuracy deteriorates leading to potential failures

Engineering Contradiction:

Improvecontrol system complexityVSAvoidengine control reliability

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent implements feedback through fault detection and accommodation mechanisms that continuously monitor inlet condition sensors. When faults are detected, the system provides feedback to switch to alternative control strategies, maintaining reliability without requiring overly complex continuous monitoring systems

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent applies beforehand cushioning by preparing alternative control strategies and fault accommodation modes in advance. When sensor faults occur, pre-prepared backup systems are activated, cushioning against the potential failures that would otherwise result from undetected sensor faults

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

Data Source

PatentUS20250320832A1Machine learned aero-thermodynamic engine inlet condition synthesis

Publication Date: 2025.10.16 RTX CORP

AI summary

A system for neural network compensated aero-thermodynamic gas turbine engine parameter/inlet condition synthesis. The system includes an aero-thermodynamic engine model configured to produce a real-time model-based estimate of engine parameters, a machine learning model configured to generate model correction errors indicating the difference between the real-time model-based estimate of engine parameters and sensed values of the engine parameters, and a comparator configured to produce residuals indicating a difference between the real-time model-based estimate of engine parameters and the sensed values of the engine parameters. The system also includes an inlet condition estimator configured to iteratively adjust an estimate of inlet conditions based on the residuals and adaptive control laws configured to produce engine control parameters for control of gas turbine engine actuators based on the inlet conditions.