Generalized approximate modeling method based on fitting sensitivity for aero-engine performance, and model application
An aero-engine and approximate model technology, which is applied in the field of generalized approximate modeling and model application of aero-engine performance based on fitting sensitivity, can solve problems such as over-fitting and under-fitting, and achieve reasonable fitting and good generalization The effect of the ability
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
specific Embodiment approach 1
[0036] Specific implementation mode one: the specific process of the generalized approximate modeling method based on the aeroengine performance of fitting sensitivity is:
[0037] Fit Sensitivity Analysis
[0038] Denote the training samples as X=[x 1 ,x 2 ,...,x n ], the fitting value is expressed as Y=[y 1 ,y 2 ,...,y n ], then when X and Y have the same initial value (x 1 =y 1 ), the degree of fit can be expressed as the fitted sensitivity model dY / dX.
[0039] (a) When dY / dX→1, y k to x k Overfitting, that is, △Y≈△X. Fitted value y k The change trend of x k consistent, such as figure 1 shown.
[0040] (b) When dY / dX>1, y k to x k Underfitting and △Y>△X, such as figure 2 shown. Fitted value y k Expanded training sample x k The trend of change, at this time y k is unstable and with x k Oscillation due to the change of , thus obtaining unstable prediction results. This state is called "over-underfitting".
[0041] (c) When 0k to x k Underfitting an...
specific Embodiment approach 2
[0068] Specific embodiment two: the difference between this embodiment and specific embodiment one is: the specific process of establishing the generalized approximate model of the aeroengine performance based on the fitting sensitivity in the step one is:
[0069] Set the value of dY / dX in the interval (0, 1), when x 1 =y 1 When building a fitted sensitivity model:
[0070]
[0071] where X=[x 1 ,x 2 ,...,x n ] is the training sample, Y=[y 1 ,y 2 ,...,y n ] is the fitting value;
[0072] (a) when|x k -y k | becomes larger, the fitted value y k Deviate from the training sample x k . because the fitted value y k Should reflect the main trend of the training sample, so the training sample x k Contains strong noise and fluctuations. in order to make y k Get the slower main trend, y k to x k The sensitivity of dy k / dx k should be lowered.
[0073] (b) When|x k -y k | becomes smaller, the fitted value y k Approximate training samples x k . because the ...
specific Embodiment approach 3
[0089] Embodiment 3: This embodiment differs from Embodiment 1 or Embodiment 2 in that the value of p is set in step 21: 1
[0090] When training the prediction model parameters, due to the large number of training samples, it is difficult to ensure that any sample segment will not be over-fitting. In order to improve the prediction accuracy, only the last p sample points of the training samples are constrained to satisfy the suppression of over-fitting. Underfitting constraints. The maximum length of p can be taken to the training sample length n, and the minimum can be taken to be 1, that is, p∈[1,n] and p∈N.
[0091] When p→1, the model only constrains the last few points, so that the fitted value of the training sample does not fall into overfitting and underfitting, but few points do not contain x k trend information, leading to inaccurate forecast results.
[0092] When p→n, the model constrains the entire training sample segment so that the fitted value of the tr...
PUM
Login to view more Abstract
Description
Claims
Application Information
Login to view more - R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap