Joint dimension reduction method based on probability box global sensitivity analysis and active subspace

Abstract

The invention discloses a joint dimension reduction method based on probability box global sensitivity analysis and active subspace, and relates to the aerospace technology. The method comprises the steps of 1) arranging uncertainty parameters of a system, 2) performing uncertainty analysis on an initial model to obtain an initial probability box, and calculating the area of the probability box; 3) reducing the uncertainty of input parameters to be analyzed, and analyzing the model to obtain the area of an output probability box after the uncertainty is reduced; 4) comparing the area change amounts of the output probability boxes before and after reduction, and calculating a global sensitivity index S of the parameter; 5) selecting another input parameter, and repeating the steps 3 and 4;6) arranging the S of each input parameter in a descending order; 7) sampling the input parameters with the uncertainty reserved, calculating model input and obtaining a gradient covariance matrix; 8)carrying out characteristic decomposition on the gradient covariance matrix, and sorting according to the size; and 9) carrying out dimensionality reduction on the input parameters by using a base vector of an active subspace. The application range is wider, and the convenience and effectiveness of subsequent modeling and optimization are effectively improved.

Description

technical field [0001] The invention relates to the field of aerospace technology, in particular to a joint dimension reduction method based on probability box global sensitivity analysis and active subspace, which is applicable to complex high-dimensional uncertainty simulation systems such as aerospace, and can effectively reduce the dimension of model input parameters . Background technique [0002] Due to various errors in manufacturing, measurement, calculation and the model itself, there are a large number of uncertain parameters in the aerospace system, including structural parameters, external forces, initial conditions and boundary conditions. Uncertainty parameters can be roughly divided into stochastic, cognitive, and mixed uncertainty parameters according to their categories. There are obvious differences in the model uncertainty caused by the mixed effect of different types of uncertainty parameters and the effect of a single uncertainty parameter. Compared wit...

AI Technical Summary

Problems solved by technology

Traditionally, when dealing with the simplification of high-dimensional parameters, only one of sensitivity analysis or dimensionality reduction methods is generally used, which is difficult to meet the needs of aerospace engineering

[0004] In order to solve the problem that the effect of reducing the dimension of input parameters by a single method is not obvious in complex engineering models, this invention combines sensitivity analysis and dimensionality reduction methods in one framework, and proposes a combination of global sensitivity analysis and active subspace based on probability box Dimensionality reduction methods for reducing the parametric complexity of the system, laying the foundation for subsequent uncertainty modeling and multidisciplinary optimization of aerospace systems

Images