Top-level approach to Simulink parameter study

I inherited a Simulink model that I will use for a study in which ~50 parameters are varied. This will require repeatedly running the model for each choice of parameter value, saving the output, and estimating summary statistics. For parameter changes that produce interesting statistics, I will need to resurrect and compare the time-varying behavior of the system. The Simulink model and choice of summary statistics will likely evolve.

Since I am new to Simulink, I would like top-level advice on how to cut into this pie. Should I have a master script mfile that walks through the parameter combinations, passes them to Simulink, and saves the result in a giant structure? Or...?

Answer 1

[Kaustubha Govind]

The SystemTest product seems exactly what you need. Check out some Demos and Webinars. You can also ask for a trial version of the product if you're not sure you want to buy it.

Alternatively, of course, you can perform similar analysis using the Simulink command-line API. I don't have much experience doing this sort of thing, but I think you would be best advised to save results from each iteration to a MAT-file and then load each MAT-file in turn to do the analysis. This will ensure that you won't lose any data for whatever untoward reason and be forced to repeat simulations.

[K E]

Thanks. For the first pass at the problem, I will run Simulink from the command line. The following tutorials appear helpful: http://www.chem.mtu.edu/~tbco/cm416/MatlabTutorialPart3.pdf , http://www.goddardconsulting.ca/simulink-interfacing-with-matlab.html .