Digital Twin Feedback for Aircraft Engine Part Quality Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial manufacturing processes often result in parts that deviate from their design intent due to manufacturing variations, particularly exacerbated in aircraft engine design where performance under high temperature conditions is affected, lacking integration of performance monitoring features.
Innovation Solution
A digital integrated process that links as-built, as-manufactured, as-designed, as-simulated, as-tested, and as-serviced components through a machine learning-based system, creating a digital twin for accurate prediction and optimization of part performance by aggregating sub-system component level predictions and updating models in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional manufacturing processes are used to produce parts according to predetermined dimensional tolerances, then manufacturing simplicity is maintained, but the as-manufactured parts deviate from design intent due to manufacturing variations
Solution Approach 1:
The system segments the manufacturing and monitoring process into distinct digital modules (design digital twin, manufacturing digital twin, service digital twin) that can be independently developed and integrated. Each digital twin captures specific aspects of the physical asset lifecycle, allowing complex manufacturing variations to be managed through modular digital representations rather than monolithic process integration.
Solution Approach 2:
The digital twin platform serves multiple functions across the entire asset lifecycle - from design validation and manufacturing monitoring to predictive maintenance and service optimization. A single digital twin infrastructure supports diverse applications including quality assessment, performance prediction, and manufacturing process improvement, eliminating the need for separate systems for each function.
2Power
If aircraft engine core components are forced to run at higher temperatures with less cooling flows, then engine power is improved, but the distribution of component robustness associated with manufacturing variations is exacerbated
Solution Approach 1:
The system implements continuous feedback loops where digital twins monitor actual component performance and manufacturing variations, then feed this information back to predictive models. This feedback mechanism allows the system to identify components with adverse manufacturing variations before they fail under high-temperature conditions, enabling proactive quality adjustments and maintaining reliability while operating at higher power levels.
Solution Approach 2:
The digital twin technology performs preliminary assessment of component robustness during the manufacturing and design phases. By predicting which components may be susceptible to high-temperature stress based on manufacturing variations, the system can pre-screen and select the most robust components for high-power applications, or adjust manufacturing parameters before production to ensure adequate robustness distribution.
3Measurement precision
If comprehensive field inspection data is collected to monitor performance conditions, then part quality assessment is improved, but data integration complexity increases
Solution Approach 1:
The system creates digital copies (digital twins) of physical assets that mirror their characteristics, performance, and history. Instead of integrating and analyzing raw data from multiple inspection sources, the digital twin serves as a unified virtual representation that already synthesizes this information, allowing quality assessment to be performed on the digital copy rather than managing the complexity of raw data integration across multiple systems.
Data Source
AI summary
There are provided methods and systems for optimizing a manufacturing process. For example, there is provided a method for generating a model for driving a decision of a manufacturing process. The method includes simultaneously receiving data from a plurality of sources and executing a machine learning-based procedure on the data. The method further includes updating a physics-based model corresponding to the model in real time based on a result of the machine learning-based procedure.