Sensitivity analysis method capable of jointly considering input parameter uncertainty and surrogate model uncertainty

Abstract

The invention discloses a sensitivity analysis method capable of jointly considering input parameter uncertainty and surrogate model uncertainty. The analysis method adopts an experiment design method to generate an initial sampling point in a definitional domain; a model is constructed through a computer, and a response value corresponding to the initial sampling point is obtained through simulation; a Kriging surrogate model is constructed by the initial sampling point and the corresponding response value; the Kriging surrogate model can carry out predication to obtain the mean value and the variance of a prediction point, at the time, the mean value and the variance can be seen as random fields, and a K-L expansion (K-L Expansion) method is adopted to convert the surrogate model uncertainty into the parameter uncertainty; and according to the concept of a subset, an initial input parameter and a K-L expansion variable are independently divided into two subsets, and a Monte Carlo method is adopted to calculate the statistical sensitivities of the two subsets. By use of the sensitivity analysis method, the influence degree of the input parameter uncertainty and the surrogate model uncertainty on the statistical sensitivities can be effectively quantized so as to guide the reasonable distribution of resources to control the influence of the input parameters on output uncertainty.

Description

technical field [0001] The invention belongs to the field of uncertainty analysis of mechanical design, in particular to a sensitivity analysis method which jointly considers the uncertainty of input parameters and the uncertainty of an agent model. Background technique [0002] Computer simulation technology (such as finite element analysis, molecular dynamics simulation) is widely used in engineering fields, such as aerospace, biomechanics, automobile manufacturing, etc. By simulating the structure, computer simulation technology can greatly improve the stiffness and strength of the product, as well as reduce weight, material and cost. However, computer simulation technology has certain limitations and problems in engineering applications, which are mainly reflected in two aspects: first, the accuracy of computer simulation results is closely related to the accuracy of model construction. However, the calculation time of high-precision computer simulation models will decre...

AI Technical Summary

Problems solved by technology

[0004] At present, the statistical sensitivity analysis based on the proxy model has received more and more attention. However, the existing methods ignore the uncertainty of the proxy model caused by the use of the proxy model, and the uncertainty of the proxy model has a great influence on the results of the statistical sensitivity analysis. will have a non-negligible impact on

Images