【灵敏度分析】应用莫里斯方法降低因素低估的风险（Matlab代码实现）

 👨‍🎓个人主页：研学社的博客 

💥💥💞💞欢迎来到本博客❤️❤️💥💥

🏆博主优势：🌞🌞🌞博客内容尽量做到思维缜密，逻辑清晰，为了方便读者。

⛳️座右铭：行百里者，半于九十。

📋📋📋本文目录如下：🎁🎁🎁

目录

💥1 概述

📚2 运行结果

🎉3 参考文献

🌈4 Matlab代码实现

---

💥1 概述

灵敏度分析用于估计不确定性因素对函数输出的影响。莫里斯方法有时被称为定性方法：它给出有限数量的计算的粗略估计。作为第一步，莫里斯方法可用于简化函数。它可以识别可以固定的低影响因素。

分析的函数必须是具有一个多维输入 x 的函数，x 必须表示分位数超空间中不确定性因子的值（x 的第 i 个坐标不是第 i 个因子的实际值，而是其累积分布函数的相应值）。要调整您的函数，首先通过将因子的累积分布函数的反函数应用于 x 来计算因子的实际值.

📚2 运行结果

部分代码：

%% 1) Clearing the memory
close all; % Closes the figures
clear all; % Clears the memory
clc; % Clears the command window

%% 2) Parameters : Please fill in

% Number of factors of uncertainty of the function studied :
nfac=20; 

% Maximum number of simulation runs :
% Large number = better estimation of the influence of the factors
% Recommended value : (number of factors + 1) * 10
% The algorithm will maybe exceed this value if it is considered necessary
nsim_max = 210;

% Function studied :
% Replace test_function by the name of your function. It must be a 
% function with one multidimensional input x. x must represent the values 
% of the uncertainty factors in the quantiles hyperspace (the i-th 
% coordinate of x is not the actual value of the i-th factor, but the 
% corresponding value of the cumulative distribution function of the i-th 
% factor). To adapt your function, first calculate the actual values of 
% the factors by applying the inverse of their cumulative distribution 
% function to each coordinate of x; Matlab includes such inverses: 
% mathworks.com/help/stats/icdf.html ).
studied_function = @(x)test_function(x);

%% 3) Initialization of the variables
table_outputs = []; % All the outputs of the simulations runs. One line = results around one point of the factors hyperspace. First column = output at a sampled point, second column = output after varying the first factor, etc...
table_ee = []; % All the elementary effects. One line = elementary effects at one point of the factors hyperspace. First column = elementary effect of the first factor, etc... See the relation between the outputs of the simulation run and the elementary effects.
factors_over = []; % Indexes of the factors over the limit (important factors). The elementary effects of these factors will not be calculated at the next step.
table_factors_over = []; % Factors over the limit at the different steps. n-th line = factors with no elementary effect at the n-th step = factors over the limit at the (n-1)-th step.
points=[]; % Sampled points of the factors hyperspace where the elementary effects are calculated.
n=1; % Current step.

🎉3 参考文献

部分理论来源于网络，如有侵权请联系删除。

[1]阚丽娟,徐吉辉,陈玉金,宋晓博.安全性要求下系统失效概率及其灵敏度分析[J].安全与环境学报,2022,22(05):2342-2351.DOI:10.13637/j.issn.1009-6094.2021.0650.

[2]Saltelli, A., Tarantola, S., Campolongo, F., and Ratto, M. (2004). Sensitivity Analysis in Practice - A Guide to Assessing Scientific Models. Wiley.

🌈4 Matlab代码实现