Usage-Based Fatigue Prediction for Gas Turbine Engines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting the cyclic life consumption of machinery components, such as gas turbine engines, are limited by their inability to account for actual operating conditions, leading to conservative lifespan estimates and potential engine damage due to unforeseen operational variations.
Innovation Solution
A method using machine learning techniques to construct models correlating low cycle fatigue consumption with flight data, allowing for real-time prediction of usage-based life consumption and remaining life of gas turbine engine components by analyzing historical and operational data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If design phase calculations and lifespan designations are used, then lifespan can be designated based on standard flights, but actual operating conditions cannot be taken into account leading to conservative estimates
Solution Approach 1:
The patent implements feedback by continuously monitoring actual operating conditions through sensors and flight data, then using this information to update and refine lifespan predictions. The system compares actual usage patterns against design phase assumptions and adjusts remaining lifespan estimates accordingly, creating a closed-loop system that learns from operational data.
Solution Approach 2:
The patent performs preliminary actions by establishing baseline lifespan predictions during the design phase using standard flight conditions, then prepares the system to receive and process actual operating data. This preliminary modeling provides an initial estimate that is subsequently refined with real-world data, allowing the system to start with conservative estimates and improve over time.
2Measurement precision
If physics-based approaches are used, then fatigue life limits can be evaluated, but computational intensity prevents near real-time application
Solution Approach 1:
The patent segments the fatigue life prediction process into two distinct phases: an offline training phase where complex physics-based models are used to establish relationships between operating parameters and fatigue consumption, and an online prediction phase where simplified machine learning models provide rapid predictions. This segmentation allows computationally intensive calculations to be performed only during training, while deployment uses lightweight models suitable for real-time application.
Solution Approach 2:
The patent employs machine learning models that serve as computationally inexpensive approximations of the full physics-based models. These ML models are trained once using extensive computational resources, then deployed as lightweight predictors that require minimal processing power during actual usage, effectively replacing expensive repeated calculations with cheap predictions.
3Reliability
If existing methods are used, then fatigue life can be predicted, but they do not include characteristics to account for actual operating conditions such as flight data
Solution Approach 1:
The patent creates a universal prediction system that can handle multiple types of operating conditions and data sources through a single machine learning model. The model is designed to process various input features including flight data, engine parameters, and environmental conditions, making it adaptable to different operational scenarios without requiring separate models for each condition type.
Solution Approach 2:
The patent utilizes parameter changes by incorporating actual operating parameters from flight data into the prediction model. The system monitors changes in engine temperature, pressure, rotational speed, and other critical parameters, then uses these dynamic parameter values to adjust lifespan predictions in real-time, reflecting the actual stress and strain conditions experienced by engine components.
Data Source
AI summary
Systems and methods for predicting usage based lifing and low cycle fatigue consumption are provided. In one example embodiment, a method can include obtaining historical flight data associated with one or more gas turbine engines of an aerial vehicle; obtaining data indicative of one or more operational conditions of the aerial vehicle during an operating period; determining whether the flight data is indicative of a usable flight; and constructing a model correlating low cycle fatigue consumption with flight data using a machine learning technique.