Geometry in Activate

Vaishnavi Kokadwar

I have done the simulation of the ball balancing table for the Altair logo. It came out to be accurate. However, I get the following warning:

Warning: Final times are not equal in MotionSolve (2) and Activate (25) models. Activate time will be used. ( 0 )

Block(s): Plant/IncludeDiagram_1_1_1/MotionSolveSignals_1/MotionSolve

Model: altair_logo

 

Also, when I tried the same simulation for a path different from the Altair logo(the Google Meet logo for example), I got a totally incorrect geometry with the ball slipping off the table completely. I got the same warning in this case as well.

I would be grateful for any help in this regard.

Vaishnavi Kokadwar

Thanks, Armin.

So, this is how I solved the geometry issue I was having earlier.

1. To get an idea of the geometry generated using Embed and Compose, we open the Activate file and double click on the Reference Trajectory block. Here, we add a ScopeXY block to take a look at the original geometry. (ScopeXY block can be found in the Palette browser within Activate-->SignalViewers.) We add the branches from xref and yref to this ScopeXY block and run the simulation. We can take a look at the original geometry by double-clicking on the ScopeXY block. For my case:

2. Depending on the software we use for extraction and your logo, the geometry might need some scaling and offsetting. In my case, I scaled the model down by 500 times and added the offset of -6000 in x and +3000 in y.

3. Scaling can be done using the Gain block in Activate-->MathOperations and inputting an appropriate value. Offsetting can be done using the Sum block in Activate-->MathOperations and a Constant signal block in Activate-->SignalGenerators. The final configuration of the system looked something like this in my case:

4. Now, we also need to change the observer settings in the Model code. There is an initial setting that is provided for the model for the case of the Altair logo. We should now change it according to our specific requirements same as we did in the reference trajectory block. The code in the modified version looks like this:

5. We can change the plant selection to Linear to get a faster result. However, this plant choice will not generate the .h3d file that is required for further processing of the Ball Balancing Table. We can also remove the 'Real-Time Scaling' as mentioned by Armin earlier for faster solving. Final Geometry obtained:

 

6. Additionally, we can also change interpolation methods within the xref and yref to explore the solution capacities further.

Also attached is the final .h3d file for the Raspberry logo.

armin_22187

Hello Vaishnavi, I'm getting the same warning.... and I ignored it so far.

Regarding your second issue: There are different possible reasons and solutions:

a) there is an offset in your logo

b) the time axis of your path is too 'aggressive'

c) your path of the google logo is rough and the controller has difficulties to follow

b) the settle time at the beginning is too short to bring the ball from the center to the start point of your path

In the same way, you have a number of possible solutions you can try. However, I want to start with a general advice:

- the calculation time of the Co-Simulation (full 3D model) is expensive - I highly recommend to select a 'simpler' representation of the BBT during the first try outs - for example the linear plant:

Open the 'Plant' in the main Window and check the linear model for example (double click on Activate linear). Later, define the parameter in the 'Model'-settings (Plant_Choice). The linear model solves more or less instantly. If you want to have the calculation done even faster, remove 'Real time scaling' in the solver settings! 

You will notice that the linear model does not work well with the highly sophisticated controller: 'PIVObserver' - The Observer requires some signals that the linear model doesn't deliver. You can check the details yourself. You may use the PID controller in the meantime: 

This will help you to see if your path doesn't contain any major hurdles.

The later step will be the smoothing of the path. You can

- increase the sampling time
- improve the spline approximation ahead of generating the .csv
- or add a low pass filter ahead of the controller (you may test some different cut off frequencies) 

A simple plot or xy plot will help you to see how well your path looks like.

Please test all the options and come back to us in case you need further recommendations.

armin

Vaishnavi Kokadwar

Thanks, Armin.

So, this is how I solved the geometry issue I was having earlier.

1. To get an idea of the geometry generated using Embed and Compose, we open the Activate file and double click on the Reference Trajectory block. Here, we add a ScopeXY block to take a look at the original geometry. (ScopeXY block can be found in the Palette browser within Activate-->SignalViewers.) We add the branches from xref and yref to this ScopeXY block and run the simulation. We can take a look at the original geometry by double-clicking on the ScopeXY block. For my case:

2. Depending on the software we use for extraction and your logo, the geometry might need some scaling and offsetting. In my case, I scaled the model down by 500 times and added the offset of -6000 in x and +3000 in y.

3. Scaling can be done using the Gain block in Activate-->MathOperations and inputting an appropriate value. Offsetting can be done using the Sum block in Activate-->MathOperations and a Constant signal block in Activate-->SignalGenerators. The final configuration of the system looked something like this in my case:

4. Now, we also need to change the observer settings in the Model code. There is an initial setting that is provided for the model for the case of the Altair logo. We should now change it according to our specific requirements same as we did in the reference trajectory block. The code in the modified version looks like this:

5. We can change the plant selection to Linear to get a faster result. However, this plant choice will not generate the .h3d file that is required for further processing of the Ball Balancing Table. We can also remove the 'Real-Time Scaling' as mentioned by Armin earlier for faster solving. Final Geometry obtained:

 

6. Additionally, we can also change interpolation methods within the xref and yref to explore the solution capacities further.

Also attached is the final .h3d file for the Raspberry logo.

Abdul Muizz Jamal

Vaishnavi Kokadwar said:

Thanks, Armin.

So, this is how I solved the geometry issue I was having earlier.

1. To get an idea of the geometry generated using Embed and Compose, we open the Activate file and double click on the Reference Trajectory block. Here, we add a ScopeXY block to take a look at the original geometry. (ScopeXY block can be found in the Palette browser within Activate-->SignalViewers.) We add the branches from xref and yref to this ScopeXY block and run the simulation. We can take a look at the original geometry by double-clicking on the ScopeXY block. For my case:

2. Depending on the software we use for extraction and your logo, the geometry might need some scaling and offsetting. In my case, I scaled the model down by 500 times and added the offset of -6000 in x and +3000 in y.

3. Scaling can be done using the Gain block in Activate-->MathOperations and inputting an appropriate value. Offsetting can be done using the Sum block in Activate-->MathOperations and a Constant signal block in Activate-->SignalGenerators. The final configuration of the system looked something like this in my case:

4. Now, we also need to change the observer settings in the Model code. There is an initial setting that is provided for the model for the case of the Altair logo. We should now change it according to our specific requirements same as we did in the reference trajectory block. The code in the modified version looks like this:

5. We can change the plant selection to Linear to get a faster result. However, this plant choice will not generate the .h3d file that is required for further processing of the Ball Balancing Table. We can also remove the 'Real-Time Scaling' as mentioned by Armin earlier for faster solving. Final Geometry obtained:

 

6. Additionally, we can also change interpolation methods within the xref and yref to explore the solution capacities further.

Also attached is the final .h3d file for the Raspberry logo.

Hey Vaishnavi!

I followed the approach step by step for the simulation of the Rasp logo but I didn't get the simulation final image as you have posted here. I got this vague result:

Please please help me with this!!!

Rahul_P1

Abdul Muizz Jamal said:

Hey Vaishnavi!

I followed the approach step by step for the simulation of the Rasp logo but I didn't get the simulation final image as you have posted here. I got this vague result:

Please please help me with this!!!

Please share screenshots of your model file and scaling and offset in the system configuration diagram (mentioned in point 3 and 4 above) 

you could also share your files here

Abdul Muizz Jamal

Rahul Ponginan_22560 said:

Please share screenshots of your model file and scaling and offset in the system configuration diagram (mentioned in point 3 and 4 above) 

you could also share your files here

here it is:

The model and final simulation:

please help me with this!!

Vaishnavi Kokadwar

Abdul Muizz Jamal said:

here it is:

The model and final simulation:

please help me with this!!

I don't see any problem with the code and the geometry. Maybe you could share the Activate file with the updated changes you've made along with your .csv file?