What is the fastest way to run same simulink model multiple times ?

If I have a large dataset say 6000 rows of data that I need to evaluate simulink model at each point

Currently I use for loop and use simout for each time to update simulink with each new value for my parameters.

I think there is more elegent approaches to do this task.

My simulink model uses components from Powerelctronics library. I assign each parameter with symbolic value that gets updated from Matlab workspace. This works fine with 6000 rows of data despite some affordable time. But more than that needs more elegent approaches.

Any suggestions will be appreciated.

Answer 1

[Jonas]

What is exactly the issue you are having with this approach? The speed, or the way you save and load data?

• If you want to run it faster, you can parallelize the simulations with the Parallel Computing Toolbox and a parfor-loop. Each instance of the simulation will run on a separate processor thread. This will give you significant speed increases.
• If data management is the issue, you may want to use the Simulink Test Toolbox to manage the parameter sweep, but I am not really sure you will benefit a lot from this since you have set up a way to run the simulations already.

I would in fact perform a parameter sweep on Simulink simulations in a very similar way as you are doing right now.

[omar khater]

Yes, My primary goal is to collect the data as fast as possible.

I wonder if there is something in simulink like to have a table and run simulations for each row in a table inherently instead of using loop to run the simulink through matlab

[Jonas]

I invite you to check out this: https://nl.mathworks.com/help/simulink/ug/running-parallel-simulations.html and this as an example: https://nl.mathworks.com/matlabcentral/answers/109190-how-can-i-use-a-parfor-loop-in-matlab-for-a-simulink-simulation-using-structs

Bottom line is basically, that you probably have a for-loop right now running each simulation. You need to create a similar loop but with 'parfor' instead of 'for', and make sure data is given to the simulation and captured out of it in a good way into variables.

[omar khater]

I have tried parfor loop but I have one problem.

In each loop iteration I need to update my parameters, I don't know why it flags in error that my parameters are not defined when calling simulink. It runs with i=7 and i=21 and then flags this error. I don't know why.

The solution here

https://www.mathworks.com/help/simulink/slref/parsim.html

reduces the time signifcantly but I have some one question

why does It run only 2 workers? Can I set it to more workers? How?

I have processor with 4 cores

[omar khater]

Another thing that I would like to try is here

https://www.mathworks.com/help/simulink/slref/rapid-accelerator-simulations-using-parsim.html

but when I open the used example sldemo_parallel_rapid_accel_sims_script

it gives me this error. I double checked that I have this compiler installed

https://www.mathworks.com/matlabcentral/fileexchange/52848-matlab-support-for-mingw-w64-c-c-compiler

ldemo_parallel_rapid_accel_sims_script
[15-Jul-2019 19:21:05] Checking for availability of parallel pool...
[15-Jul-2019 19:21:05] Loading Simulink on parallel workers...
Analyzing and transferring files to the workers ...done.
[15-Jul-2019 19:21:07] Configuring simulation cache folder on parallel workers...
[15-Jul-2019 19:21:07] Running SetupFcn on parallel workers...
[15-Jul-2019 19:21:11] Cleaning up parallel workers...
Error using sldemo_parallel_rapid_accel_sims_script (line
121)
Error executing SetupFcn

Caused by:
    Error using build_rapid_accel_target
    Failed to build model 'sldemo_raccel_engine_idle_speed'
    using LCC in rapid accelerator mode, please consider
    using a different compiler instead. For details on
    supported compiler versions, see

    http://www.mathworks.com/support/compilers/current_release
            Error using slbuild_private
        It appears as though pwd (H:\MatlabScripts) contains
        an existing build folder, while the root build folder
        in Simulink.fileGenControl is set to another folder.
        The build folder in pwd may shadow build artifacts in
        the root build folder and cause unexpected errors.
        Consider CDing to a folder that does not contain an
        existing build folder prior to initiating a build.
        The Build Process will terminate.
    Error using build_rapid_accel_target
    Failed to build model 'sldemo_raccel_engine_idle_speed'
    using LCC in rapid accelerator mode, please consider
    using a different compiler instead. For details on
    supported compiler versions, see

    http://www.mathworks.com/support/compilers/current_release
            Error using slbuild_private
        It appears as though pwd (H:\MatlabScripts) contains
        an existing build folder, while the root build folder
        in Simulink.fileGenControl is set to another folder.
        The build folder in pwd may shadow build artifacts in
        the root build folder and cause unexpected errors.
        Consider CDing to a folder that does not contain an
        existing build folder prior to initiating a build.
        The Build Process will terminate.