Aeroengine Parameter Normalization via Segmented Regression Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for normalizing aeroengine operating parameters struggle to account for variations due to external context variables, leading to inaccuracies in comparing engine behavior across different conditions, especially between engines of different ages or states.
Innovation Solution
A normalization method involving two regression models is employed, where the first model accounts for external context variables and the second model is specific to predefined value vector classes, allowing for more accurate normalization that considers engine-specific features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional normalization methods are used to standardize operating parameter values, then the values become comparable across different conditions, but the method fails to adequately account for dependencies on exogenous variables and engine-specific characteristics
Solution Approach 1:
The patent divides the normalization process into two distinct stages: a first normalization step using a generic regression model to handle exogenous variables, and a second normalization step using engine-specific regression models. This segmentation allows each stage to address different aspects of the normalization problem, improving overall accuracy while maintaining adaptability to engine-specific characteristics.
Solution Approach 2:
The patent performs preliminary normalization using a generic model before applying engine-specific adjustments. The first normalization step removes the influence of exogenous variables from operating parameters, creating a baseline that is then refined in the second step using engine-specific models. This preliminary action ensures that subsequent engine-specific analysis is not confounded by external conditions.
2Ease of manufacture
If a single generic regression model is used for normalization, then the method is simple to implement, but it cannot capture engine-specific variations and aging effects
Solution Approach 1:
The patent segments the regression modeling into two levels: a generic model applicable to all engines of a type, and engine-specific models tailored to individual engines. The generic model provides a standardized baseline that is simple to implement, while the engine-specific models capture individual variations and aging effects, thereby resolving the contradiction between implementation simplicity and normalization accuracy.
Solution Approach 2:
The patent applies different levels of model specificity to different normalization needs. The generic regression model provides a standardized approach for handling exogenous variables across all engines, while engine-specific regression models provide localized adjustments for individual engine characteristics. This local quality approach allows the system to maintain simplicity where applicable while achieving precision where needed.
3Measurement precision
If multiple regression models are used to account for engine-specific features, then normalization accuracy improves, but the device complexity and computational requirements increase
Solution Approach 1:
The patent segments the modeling complexity into two distinct phases: building engine-specific regression models during an offline training phase using historical data, and applying these pre-built models during the online normalization phase. This segmentation allows the system to achieve high normalization accuracy through multiple specialized models while keeping the operational complexity low, as the models are prepared in advance and simply applied during monitoring.
Solution Approach 2:
The patent performs the complex model-building activity in advance during an offline training phase, using historical operating data to construct engine-specific regression models. These pre-construction models are then stored and applied during normal operation. This preliminary action transfers computational complexity from the operational phase to the preparation phase, reducing real-time device complexity while maintaining high normalization accuracy.
Data Source
AI summary
A normalization method includes: normalizing a current value of each operating parameter relative to exogenous variables using a first regression model defined on a space generated by the exogenous variables; associating the vector formed by the normalized current values of the parameters with at least one vector class of a set of predefined classes; using at one second regression model defined on the space generated by the exogenous variables for the at least one class associated with the vector to construct, for each parameter, a third regression model onto the space; and normalizing the normalized current value of each parameter relative to the exogenous variables using the third regression model.