Gradient-enhanced variable-fidelity proxy model modeling method

A technology of proxy model and modeling method, applied in the field of proxy model, can solve the problems of reducing modeling efficiency, increasing the dimension of correlation matrix, and not fully considering the physical meaning, so as to achieve the effect of strengthening prediction accuracy and ensuring calculation efficiency

CN113868853APending Publication Date: 2021-12-31DALIAN UNIV OF TECH
0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Current Assignee / Owner
Publication Date
2021-12-31

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses a gradient-enhanced variable-fidelity proxy model modeling method, and belongs to the technical field of proxy models. The method comprises the following steps: (1) establishing a low-fidelity model according to a low-fidelity sample and a low-fidelity real response and a low-fidelity real gradient corresponding to the low-fidelity sample; taking a gradient-enhanced radial basis function model as a low-fidelity model, and giving a low-fidelity prediction response and a low-fidelity prediction gradient at a high-fidelity sample through the established low-fidelity model; (2) introducing a scale factor and a difference function to correct the low-fidelity model; (3) solving a scale factor and an undetermined coefficient column vector in the difference function; and (4) establishing a prediction expression of the gradient-enhanced variable fidelity proxy model. According to the method, the physical significance of gradient information can be fully considered on the basis of a traditional variable fidelity proxy model, so that the prediction precision of the model is further enhanced. Meanwhile, Cholesky decomposition is used when inversion is carried out on the established correlation matrix, and the calculation efficiency can be effectively guaranteed.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention belongs to the technical field of proxy models, and relates to a modeling method of a gradient-enhanced variable-fidelity proxy model. Background technique

[0002] The surrogate model establishes a mathematical approximation for the relationship between the design variables and the response of the complex system, focusing on solving the problems of excessive calculation in engineering design or the inability of existing computing resources to meet the computing needs. It has been used in different Scientific fields (such as design optimization, material design, reliability analysis, digital twins, and uncertainty quantification, etc.) have received extensive attention and applications. In addition, surrogate models can be of great help in dealing with noisy or missing data and in gaining insight into complex functional relationships between design variables and output responses.

[0003] Although the surrogate model method saves a lot o...

Examples

Embodiment Construction

[0016] The present invention will be further elaborated below in conjunction with the accompanying drawings and specific embodiments.

[0017] In order not to lose generality, there are n high-fidelity samples is the high-fidelity real response and high-fidelity real gradient corresponding to the i-th high-fidelity sample, where i=1, 2, ..., n; there are p low-fidelity samples x L ={x l1 , x l2 ,...,x lp}, is the high-fidelity true response and high-fidelity true gradient corresponding to the j-th low-fidelity sample. x hi (i=1, 2, ..., n) and x lj (j=1, 2, . . . , p) has a dimension of d. A gradient-enhanced variable-fidelity proxy model modeling method designed by the present invention, such as figure 1 As shown, the method includes the following steps:

[0018] (1) Build a low-fidelity model based on low-fidelity samples and their corresponding low-fidelity ground truth responses and low-fidelity ground truth gradients. The present invention uses the gradient-en...