Error Function Optimization for Gas Turbine Performance Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for analyzing engine performance, particularly in gas turbines, face challenges in accurately identifying performance-degrading components due to complex systems and limitations in existing analysis tools, such as ANSYN and COMPASS, which struggle with instrumentation errors and faults, and fail to provide exact solutions when the number of variables exceeds measurements.
Innovation Solution
A method that groups system variables to account for known mechanisms affecting performance, allowing independent and dependent adjustments, using a sum of absolute variations error function optimized with a least-absolutes approach, and incorporating composite variables to focus on specific performance-changing mechanisms, thereby reducing 'smearing' and improving fault identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the number of system variables is increased to cover more performance parameters, then the comprehensiveness of performance analysis is improved, but the ability to obtain exact solutions deteriorates when variables exceed measurements
Solution Approach 1:
The patent segments system variables into two distinct groups: those adjusted independently and those adjusted dependently. This segmentation allows the analysis to handle more variables than measurements by structuring the optimization problem so that dependent variables are constrained by predetermined ratios, reducing the effective degrees of freedom and enabling exact solutions even when total variables exceed measurements.
2Reliability
If least squares optimisation is used to distribute errors across multiple variables, then the statistical validity is improved, but the identification of specific faulty components deteriorates due to smearing
Solution Approach 1:
The patent segments the error distribution mechanism by applying different adjustment strategies to different variable groups. Independent variables can concentrate errors to identify specific faults, while dependent variables maintain statistical relationships. This segmentation resolves the contradiction by allowing both statistical validity and precise fault identification to coexist.
Solution Approach 2:
The patent applies local quality by allowing different portions of the system variables to have different adjustment characteristics. The first portion (independent variables) allows error concentration for precise fault identification, while the second portion (dependent variables) maintains predetermined ratios for statistical validity. This local differentiation enables both objectives to be achieved in their respective domains.
3Adaptability or versatility
If all system variables are adjusted independently to minimise error, then the optimisation flexibility is improved, but the physical consistency of performance relationships deteriorates
Solution Approach 1:
The patent segments variables into independent and dependent groups, where independent variables provide optimization flexibility and dependent variables maintain physical consistency through predetermined ratios. This segmentation allows the system to achieve both flexibility in error minimization and stability in physical relationships.
Solution Approach 2:
The patent changes the parameter adjustment strategy by applying different rules to different variable groups. Independent variables are adjusted freely to minimize error, while dependent variables are adjusted maintaining predetermined ratios that reflect physical relationships. This parameter differentiation resolves the contradiction between flexibility and consistency.
Data Source
Figure 1~2b
Figure 3
AI summary
A method is provided of analysing measured parameter values, in relation to corresponding expected parameter values, from a system which has a plurality of associated system variables. The method includes the step of optimising an error function which relates said measured parameter values to the system variables. The step of optimising the error function includes adjusting the system variables to vary the expected parameter values and thereby vary the amount of error represented by the error function. A portion of the system variables are adjusted independently of each other. A further portion of the system variables are adjusted dependently of each other. In the further portion, the system variables remain in a predetermined constant ratio relative to each other, the predetermined constant ratio being characteristic of a known mechanism changing the performance of the system.