Cosimulation, Motion view/Motion solve and Simulink/Matlab

Hello,

I am working on a cosimulation between simulink and Motion view/ moton solve right now.

But when I start the cosimulation has convergence Problems that where shown in the log file of motion view.

I checked the debug mode and found the errors that say that there are problems with the MEX-File and the libraries, there was also writen that the icudt61.dll file couln't be used. I searched the web and found in the forum that the .dll file might only be for the 32 bit version of matlab but not for the 64bit (the version I have) here is the link to the forum page I found: https://www.mathworks.com/matlabcentral/answers/488651-need-icudt61-dll-when-using-vlfeat-in-matlab-2019a

I wonderd If any body had that problem before, ich this could be the reason for the convergence problem and how I could fix it.

thanks in advance and kind regards

Jasmin Heyser

Find more posts tagged with

English

MotionView/MotionSolve

Accepted answers

All comments

Chris Coker_21312

Hi Jasmin-

One initial suggestion, make sure the MotionSolve model solves first without cosimulation.

Also, you may want to try one of the co-simulation example/tutorial models (if you haven't already), just to make sure the workflow and software setup work (this would eliminate any potential issues with your modeling setup)

Best,

Chris

Jasmin Heyser

Hi Chris,

I checked that the MotionSolve model solves without the cosimulation and the tutorials for co-simulation worked fine.

Is it possible that the Plant_Input variables of the motions could be the cause? that they can't be differentiated two times?

best regards

Jasmin

Chris Coker_21312

I don't see a reason why you can't differentiate two times. However, with some mechanical systems, 2nd derivatives can become noisy, or very discontinuous. This may be the source of the issue.

Chris

Jasmin Heyser

The cosimulation does't run though. I tried to run the motionsolve model without the cosimultion to make sure that the model is fine. I used curves as Inputs and It worked in pure MS simulation because the curve points are interpolated with cubic polynomials that can form first and second time derivatives. However the convergence problem is still there. Could it be that the problem is caused by the use of Angle sizes as Plant_Inputs for which no derivation information is available?

Jasmin

Chris Coker_21312

I guess it's possible. At this point, there appears to be something specific in your model that's causing the issue. I would take a step back, and try a simpler co-simulation setup. Without seeing the model, and understanding the specific details, it's really hard to make useful suggestions.

One thing you can do while testing your model, is to create outputs that report what the solver variables are calculating what is being passed through cosimulation. That way you can verify that what you expect to be passed to the cosimulaton has the correct scale, etc.

Chris