Most relevant variables to a Simulink model

2 views (last 30 days)
peetman
peetman on 30 Mar 2022
Answered: Michael on 24 May 2022
Hi there,
I have a complex Simulink model which controls a powertrain. There are several variables that impact the outputs. I would like to identify which variables have the most influence in a specific output.
For example, I want to evaluate the variables that have more influence in the heat rejection from my model. I would imagine a tool where I can select different variables, set upper/lower limits and step size, and the tool would run iterations by variyng the variables values.
Is there some default tool in Simulink for such analysis?
Thanks
  1 Comment
Mathieu NOE
Mathieu NOE on 30 Mar 2022
hello
why not make a for loop to test output sensivity for each parameter ?
you run sim('mymodel.slx') inside the for loop and extract the required output and get a sensivity graph from there

Sign in to comment.

Answers (2)

Paul
Paul on 30 Mar 2022
The Simulink Design Optimization toolbox can do sensitivity analysis. I've never used it, so can't say anything more than that.

Michael
Michael on 24 May 2022
You can use Sensitivity Analyzer app which is part of Simulink Design Optimization.
Here's an example that determines the influential parameters in the context of doing parameter estimation. The workflow (especially the sensitivity analysis part) is similar for response optimization.
Note that the sensitivity analysis in this app is global SA, i.e. the workflow leans towards varying all the parameters, not one-at-a-time. The parameters can be varied in a random way (pseudo-random, or in many cases quasi-random works even better to sample the parameter space, and is available if you have Statistics and Machine Learning Toolbox). Or you can variy the parameters in a gridded way.
The main 3 steps in sensitivity analysis are (1) Sample the parameter space, (2) Evaluate the model at those samples, i.e. parameter combinations, and (3) Analyze the evaluation results to determine which parameters have the most influence on the cost function, e.g. making a "tornado plot".

Products


Release

R2018b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!