Gas Turbine Compressor Operability Using ML Damage Surrogates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting compressor operability in gas turbine engines are time-consuming, labor-intensive, and not sufficiently accurate, leading to potential unnecessary condemnation of components and increased costs due to inaccurate damage assessments and complex repair requirements.
Innovation Solution
A computer-implemented method using machine learning algorithms trained with data quantifying damage to compressor blades, which includes inputting damage data and operating parameters to determine the operability of a second compressor, employing techniques such as Fourier-related transforms and artificial neural networks for accurate prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If computational fluid dynamic (CFD) techniques are used to model the effect of damage on compressor operability, then the prediction accuracy may be improved, but the time consumption and labor intensity increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning algorithms using CFD simulation data before actual damage assessment is needed. The system performs CFD simulations in advance to generate training datasets, then uses these pre-computed results to train ML models that can quickly predict compressor operability without requiring real-time CFD calculations. This resolves the contradiction by shifting computational burden from the prediction phase to the training phase.
Solution Approach 2:
The patent uses copying by creating simplified surrogate models (machine learning algorithms) that replicate the behavior of complex CFD simulations. Instead of running full CFD analyses for each damage assessment, the system uses trained ML models that copy the essential predictive capabilities of CFD at a fraction of the computational cost and time, while maintaining sufficient accuracy for operational decisions.
2Productivity
If accurate predictions of blade damage impact are made quickly, then component condemnation can be optimized, but the complexity of damage assessment increases
Solution Approach 1:
The patent applies mechanics substitution by replacing complex manual damage assessment processes with automated machine learning algorithms. The system substitutes expert human analysis and complex engineering judgment with trained ML models that automatically process damage data and predict compressor operability, thereby increasing assessment speed while managing complexity through algorithmic automation rather than human expertise.
Solution Approach 2:
The patent uses parameter changes by transforming complex damage geometry data into simplified numerical parameters that feed into machine learning models. Instead of analyzing detailed three-dimensional damage geometries directly, the system extracts key parameters (such as damage volume, location, and severity metrics) and uses these transformed parameters for rapid prediction, balancing accuracy with computational efficiency.
3Reliability
If CFD techniques are used for damage assessment, then detailed analysis may be achieved, but the process requires days of expert time and is not sufficiently accurate
Solution Approach 1:
The patent applies self-service by creating an automated assessment system that performs damage evaluation without requiring expert human intervention. The machine learning models are trained to independently assess compressor operability based on input damage data, eliminating the need for days of expert analysis while maintaining reliability. The system serves itself by automatically processing data, making predictions, and providing results without human oversight.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method comprising: controlling input of data quantifying damage received by a compressor of a gas turbine engine into a first machine learning algorithm; receiving data quantifying a first operating parameter of the compressor as an output of the first machine learning algorithm; and determining operability of the compressor by comparing the received data quantifying the first operating parameter of the compressor with a threshold.