How to change size of simscape multibody models

When I changed the size of the Stewart Platform model to 10 times the default size, the model fell apart and failed as the figure shows. What other model paprameters do I need to modify after changing the size?

The attachment is the original model.

Thank you for your help!

[Sandeep Mishra]

Hi @Ting

Please upload your model so the community can assist you more effectively.

[Ting]

Hi@Sandeep Mishra

Thank you for your attention!

I already upload the original Stewart platform model.I'll very appreciate it if you could give me some guidance.

[Yifeng Tang]

Hi @Ting,

The model you uploaded seems to be the original one from Simscape Multibody documentation. Could you please let the community know what parameters you attempted to change and ran into error? Some screenshots and explanation will be helpful.

Answer 1

[Yifeng Tang]

Hi @Ting,

I just multiplied most of the dimensions in the subsystem mask by 10, and the model seems to assemble OK. See below. Are you able to reproduce this?

[Ting]

Hi@Yifeng Tang,

Thanks for your attention but Im sorry I can't reproduce it with tR2023a.

My parameters setting is same as yours but the model doesn't work properly when I click Run,

[Yifeng Tang]

Looks like it's assembled correctly at the very beginning, all the joints appear properly constrained throughout the simulation, but the platform wasn't properly actuated and controlled.

I suspect it's a control problem. As the whole dimension is scaled up by 10, the weight is probably 1000 times heavier. The force from the original controller is probably not sufficient to support the mechanism. I suggest that you look into that.

Another thought is to make this *10 a variable. Start with maybe 1.5 times bigger. See if the controller still works and if the gains need to be larger. This may help you pinpoint the issue better.

[Ting]

Thank you for your answer!@Yifeng Tang

When I changed the model to 1.5 times bigger, changing the control gains had little effect. Although the upper platform of the model gradually returned to the normal position as the gains increased, it could never return to the correct position. After the controller gains were adjusted to a large value, the solver crashed and it was almost impossible to run the simulation. I am thinking about whether there are other control methods suitable for this model. The PID controller may not be suitable. Do you have any insights? Looking forward to your answer!

[Yifeng Tang]

When the dimensions are now 1.5x of what it was, the control reference/target probably needs to be different as well. Did you scale that up? In this particular case, the dz reference needs to be scaled up. This seems to give me nearly identical looking results as before.

Then, you can start tuning the controller gain. When I tried to scale by 2.5, the original control gains are no longer sufficient. Using reasonably higher values will stablize the platform. Repeat the steps and I was able to bring it to 10x larger. I attached the model and the script with scale values and sample control gains. You can uncomment part of the param_scale.mlx script to see the progression towards 10x larger.

% scale = 1.0;
% Kp = 300;
% Ki =  20;
% Kd =   5;

% scale = 1.5;
% Kp = 300;
% Ki =  20;
% Kd =   5;

% scale = 2.5;
% Kp = 1000;
% Ki =  100;
% Kd =   20;

% scale = 5.0;
% Kp = 4000;
% Ki =  400;
% Kd =  100;

scale = 10;
Kp = 20000;
Ki =  2000;
Kd =   500;

BTW: crazily high control gains almost always lead to unstable system. Take small steps and happy modeling.

[Ting]

Thanks a lot!@Yifeng Tang

Following your instructions, I was able to implement it successfully. These parameters do need to be changed uniformly, which I had overlooked before. Thank you very much for your continued guidance.Wish you all the best.