Can I see intermediate results durint the solution?

Yoseob Shim

Hello,

During the solving period, MotionSolve does not create h3d file. It makes h3d file when the solution is completed.

My solution is usually lasted one or two days and I want to check intermediate status if it goes well. to do check the status, I am using simple script (mspost.exe is executed every 30 minutes, repeatedly).

Is there any option or command that allow MotionSolve to write h3d file during solution automatically?

 

And one more..

I am doing many parametric studies so far and thus each models are somewhat different such as geometries or parameters. When I start new parameter study, I want to save all geometries and parameters on new folder. To do this, I am using "File/Export/Model" but some geometries are still linked to the folder defined in previous model. 

Could you let me know the way that I can save all geometries and parameters in new folder, perfectly?

 

Thank you in advance.

 

Regards,

Yoseob.

GTT Adam

Yoseob Shim said:

Hello Adam,

Thank you for your answers.

I studied link that you introduced, thank you for that. And accoding to your mention, I decided to use mspost to keep current solving time.

Regarding solving time of 1 or 2 days, I am solving model described in below picture. The machine is moving forward by 1m/s and it should pass several blocks in its forward (marked by yellow circle). And there are many contacts between roller, sprocket, idler, track shoe and ground. In addition, there is some control function to make the machine to move straight forward. I guess this is the one of major reason of long time solution but I am very appreciated if you suggest any advice to reduce solving time. Actually I was structure and durability analyst and I started dynamics and MotionSolve study this May. So any recommendation and advices are welcomed and very helpful to me.

 

Regards,

Yoseob.

 

Hi Yoseob,

In that case welcome to the world of MBD! Looks like you're well on your way to getting the most out of MotionView/MotionSolve.

A general approach for these longer types of simulations is to reduce the model size as must as you can. Depending on the point of this study, if there are any bodies that can be removed or merged it would help to reduce the degrees of freedom of the system and reduce the simulation time.

For example, if the point of this simulation is to measure the force at the cab rotation bearing, then all of the parts of the cab, boom, and bucket can all be combined into a single body with an asserted mass/cog/inertia of the assembly. This could reduce the number of components and degrees of freedom significantly. Also, high fidelity geometry like parasolids can increase solver time as opposed to simple shapes. Once again, if the point is to study the cab rotational bearing, then I would suppress the CAD for the cab, boom, and bucket and replacing it with a simple cube with asserted properties.

If the model complexity can't be reduced, then you may want to reduce the number of obstacles you encounter. Again, depending on the goal of this simulation, you may only need to hit 1 left obstacle and 1 right obstacle instead of the 4 each shown in the picture.

The last place I would also look into would be your solver settings. If error tolerance and maximum integrator step sizes are set drastically low, this will significantly slow down solving time.

 

Hope this helps!

Adam Reid

Ganesh Shanmugam

Hello Yoseob,

The H3d will be generated after the simulation. As you are doing, using mspost is the only option to get the H3d during the simulation.

In File --> Export Model use Archive Model option to save all geometries and parameters in new folder.

Regards,
S. Ganesh

GTT Adam

Hi Yoseob,

Another way to view the animation during the solution is to run "Online" using the newer interface of MotionView. This will display the animation as it progresses, however the overall solution time may increase due to pushing animations to your active session while solving. 

You mention a solution time of 1 or 2 days, is that for an entire design study or for a single simulation? If this pertains to a single simulation, it may be worthwhile to reduce the complexity of the model or the number of degrees of freedom. One of the strengths of multibody dynamics is the ability to simulate highly transient models quickly, and long runtimes are usually remedied by improving the modeling approach or by reducing the number of computationally expensive entities like FlexBodies. 

 

Hope this helps!

Adam Reid

Praful

Hi Yoseob - Another approach that you may try is to write a Model Animation File (MAF) (there is a checkbox) and load combination of maf + mrf in HyperView.

Note, this may not support all types of graphics or flexible bodies.

If you use the Archive functionality, then it should copy all geometries along with the model to the new folder/zip.

 

Praful

Yoseob Shim

Adam Reid_21142 said:

Hi Yoseob,

Another way to view the animation during the solution is to run "Online" using the newer interface of MotionView. This will display the animation as it progresses, however the overall solution time may increase due to pushing animations to your active session while solving. 

You mention a solution time of 1 or 2 days, is that for an entire design study or for a single simulation? If this pertains to a single simulation, it may be worthwhile to reduce the complexity of the model or the number of degrees of freedom. One of the strengths of multibody dynamics is the ability to simulate highly transient models quickly, and long runtimes are usually remedied by improving the modeling approach or by reducing the number of computationally expensive entities like FlexBodies. 

 

Hope this helps!

Adam Reid

Hello Adam,

Thank you for your answers.

I studied link that you introduced, thank you for that. And accoding to your mention, I decided to use mspost to keep current solving time.

Regarding solving time of 1 or 2 days, I am solving model described in below picture. The machine is moving forward by 1m/s and it should pass several blocks in its forward (marked by yellow circle). And there are many contacts between roller, sprocket, idler, track shoe and ground. In addition, there is some control function to make the machine to move straight forward. I guess this is the one of major reason of long time solution but I am very appreciated if you suggest any advice to reduce solving time. Actually I was structure and durability analyst and I started dynamics and MotionSolve study this May. So any recommendation and advices are welcomed and very helpful to me.

 

Regards,

Yoseob.

 

Yoseob Shim

Ganesh Shanmugam said:

Hello Yoseob,

The H3d will be generated after the simulation. As you are doing, using mspost is the only option to get the H3d during the simulation.

In File --> Export Model use Archive Model option to save all geometries and parameters in new folder.

Regards,
S. Ganesh

Hello Ganesh.

Thank you for the answers.

 

Regards,

Yoseob.

Yoseob Shim

Praful Prabhu_20784 said:

Hi Yoseob - Another approach that you may try is to write a Model Animation File (MAF) (there is a checkbox) and load combination of maf + mrf in HyperView.

Note, this may not support all types of graphics or flexible bodies.

If you use the Archive functionality, then it should copy all geometries along with the model to the new folder/zip.

 

Praful

Hello Praful,

Thank you for your answer.

I will try with your advice in following solving.

 

Regards,

Yoseob.

GTT Adam

Yoseob Shim said:

Hello Adam,

Thank you for your answers.

I studied link that you introduced, thank you for that. And accoding to your mention, I decided to use mspost to keep current solving time.

Regarding solving time of 1 or 2 days, I am solving model described in below picture. The machine is moving forward by 1m/s and it should pass several blocks in its forward (marked by yellow circle). And there are many contacts between roller, sprocket, idler, track shoe and ground. In addition, there is some control function to make the machine to move straight forward. I guess this is the one of major reason of long time solution but I am very appreciated if you suggest any advice to reduce solving time. Actually I was structure and durability analyst and I started dynamics and MotionSolve study this May. So any recommendation and advices are welcomed and very helpful to me.

 

Regards,

Yoseob.

 

Hi Yoseob,

In that case welcome to the world of MBD! Looks like you're well on your way to getting the most out of MotionView/MotionSolve.

A general approach for these longer types of simulations is to reduce the model size as must as you can. Depending on the point of this study, if there are any bodies that can be removed or merged it would help to reduce the degrees of freedom of the system and reduce the simulation time.

For example, if the point of this simulation is to measure the force at the cab rotation bearing, then all of the parts of the cab, boom, and bucket can all be combined into a single body with an asserted mass/cog/inertia of the assembly. This could reduce the number of components and degrees of freedom significantly. Also, high fidelity geometry like parasolids can increase solver time as opposed to simple shapes. Once again, if the point is to study the cab rotational bearing, then I would suppress the CAD for the cab, boom, and bucket and replacing it with a simple cube with asserted properties.

If the model complexity can't be reduced, then you may want to reduce the number of obstacles you encounter. Again, depending on the goal of this simulation, you may only need to hit 1 left obstacle and 1 right obstacle instead of the 4 each shown in the picture.

The last place I would also look into would be your solver settings. If error tolerance and maximum integrator step sizes are set drastically low, this will significantly slow down solving time.

 

Hope this helps!

Adam Reid

Yoseob Shim

Adam Reid_21142 said:

Hi Yoseob,

In that case welcome to the world of MBD! Looks like you're well on your way to getting the most out of MotionView/MotionSolve.

A general approach for these longer types of simulations is to reduce the model size as must as you can. Depending on the point of this study, if there are any bodies that can be removed or merged it would help to reduce the degrees of freedom of the system and reduce the simulation time.

For example, if the point of this simulation is to measure the force at the cab rotation bearing, then all of the parts of the cab, boom, and bucket can all be combined into a single body with an asserted mass/cog/inertia of the assembly. This could reduce the number of components and degrees of freedom significantly. Also, high fidelity geometry like parasolids can increase solver time as opposed to simple shapes. Once again, if the point is to study the cab rotational bearing, then I would suppress the CAD for the cab, boom, and bucket and replacing it with a simple cube with asserted properties.

If the model complexity can't be reduced, then you may want to reduce the number of obstacles you encounter. Again, depending on the goal of this simulation, you may only need to hit 1 left obstacle and 1 right obstacle instead of the 4 each shown in the picture.

The last place I would also look into would be your solver settings. If error tolerance and maximum integrator step sizes are set drastically low, this will significantly slow down solving time.

 

Hope this helps!

Adam Reid

Hello Adam,

I really appreciate for your impressive advices.

Actually I didn't thought about simplification (replacing to mass with inertia) of non-interesting parts, I need to keep your advice in mind.

In addition to your advice, my track/roller/sprocket/idler/ground have 5mm mesh size to get accurate response under the contact situation. But I guess that there would be no problem although I mesh them with larger mesh size(10~20mm). In stress analysis, a lot of solving time is spent to calculate contact searching and treating. How about MotionSolve? Is it spend lots of time too? Please let me know your experience and knowledge if the contact is trivial factor in MotionSolve run time.

For my half years study, I am learning many things and in next, I guess I can make better MotionSolve model by reflecting my eperiences and your impressive advices. I appreciated for that, again.

 

Best Regards,

Yoseob.

 

GTT Adam

Yoseob Shim said:

Hello Adam,

I really appreciate for your impressive advices.

Actually I didn't thought about simplification (replacing to mass with inertia) of non-interesting parts, I need to keep your advice in mind.

In addition to your advice, my track/roller/sprocket/idler/ground have 5mm mesh size to get accurate response under the contact situation. But I guess that there would be no problem although I mesh them with larger mesh size(10~20mm). In stress analysis, a lot of solving time is spent to calculate contact searching and treating. How about MotionSolve? Is it spend lots of time too? Please let me know your experience and knowledge if the contact is trivial factor in MotionSolve run time.

For my half years study, I am learning many things and in next, I guess I can make better MotionSolve model by reflecting my eperiences and your impressive advices. I appreciated for that, again.

 

Best Regards,

Yoseob.

 

Hi Yoseob,

In general, contacts are more computationally expensive than constraints, so if a contact is not needed and can be replaced with a joint then it is recommended.

Depending on the type of contact, it is recommended to reduce '3D to 3D' contacts to something like a '2D point to curve' instead as this would also speed up solver time if possible.

As well, if '3D to 3D' contact is required, then you can always partition the contact bodies to make the contact searching easier on the solver. For example, if we were simulating a gear-train with contacts, you can partition a gear in HyperMesh such that the teeth are individual bodies and then create multiple contacts. This way, the solver is not checking to see if every single node on the gear is in contact or not at every integration step.

 

Hope this helps!

Adam Reid

Yoseob Shim

Adam Reid_21142 said:

Hi Yoseob,

In general, contacts are more computationally expensive than constraints, so if a contact is not needed and can be replaced with a joint then it is recommended.

Depending on the type of contact, it is recommended to reduce '3D to 3D' contacts to something like a '2D point to curve' instead as this would also speed up solver time if possible.

As well, if '3D to 3D' contact is required, then you can always partition the contact bodies to make the contact searching easier on the solver. For example, if we were simulating a gear-train with contacts, you can partition a gear in HyperMesh such that the teeth are individual bodies and then create multiple contacts. This way, the solver is not checking to see if every single node on the gear is in contact or not at every integration step.

 

Hope this helps!

Adam Reid

Good morning Adam,

Thank you for your answers.

According to your answer, I understand that contact in MotionSolve is also important factor for solving time. I need to be careful in modeling as your advice.

Actually, I am familiar with contacts not only Penalty Method but also Lagrangian Method because these are frequently used in my stress analyses as well. But I have few experience in 2D contact and I need to be familiar with that because I am in Dynamics simulation region now. I need to modify the way of thinging to be aligned to Dynamics.

 

I really appreciate for your warm and kind advice everytime.

 

Best Regards,

Yoseob.

GTT Adam

Yoseob Shim said:

Good morning Adam,

Thank you for your answers.

According to your answer, I understand that contact in MotionSolve is also important factor for solving time. I need to be careful in modeling as your advice.

Actually, I am familiar with contacts not only Penalty Method but also Lagrangian Method because these are frequently used in my stress analyses as well. But I have few experience in 2D contact and I need to be familiar with that because I am in Dynamics simulation region now. I need to modify the way of thinging to be aligned to Dynamics.

 

I really appreciate for your warm and kind advice everytime.

 

Best Regards,

Yoseob.

Hi Yoseob,

Not a problem. I'm glad I could provide some guidance. With your background and with a few of these tips, you should be able to cut down the solver time of your model dramatically.

If you feel one of my responses answered your original question, kindly mark that response as an accepted answer for other readers on this forum.

 

Thanks,

Adam Reid

Yoseob Shim

Adam Reid_21142 said:

Hi Yoseob,

Not a problem. I'm glad I could provide some guidance. With your background and with a few of these tips, you should be able to cut down the solver time of your model dramatically.

If you feel one of my responses answered your original question, kindly mark that response as an accepted answer for other readers on this forum.

 

Thanks,

Adam Reid

Hello Adam,

Your replies were fully answered on my question so I clicked on the checkbox. Please let me know if I did incorrectly.

Thank you again and have a good weekend.

 

Best Regards,

Yoseob.