Hello,
I have compared the thelen2003 and millard2012 muscle models by plotting the default values for the knee flexion moment against the knee flexion angle for the knee flexors and extensors in my model and I am really surprised to find that the Millard2012 has really small moments in comparison to the Thelen model (6-8 x smaller). Is there a reason behind this ?
Because I need to strengthen my model and with the thelen I can do 3x the max isometric force but with Millard I would have to do 20x the max isometric force to get the value I want and then there is too many problem due to the enormous passive forces.
Thanks,
Frederic
moment vs angle plot for thelen and millard model
-
Frédéric Lamon
- Posts: 27
- Joined: Wed May 06, 2015 1:01 pm
-
Matthew Millard
- Posts: 24
- Joined: Fri Aug 01, 2008 7:17 am
Re: moment vs angle plot for thelen and millard model
Hello Frédéric,
Have you checked the fiber lengths and the value of the active-force-length multiplier? My guess is that the normalized fiber length is less than 0.6 or greater than 1.5. In these regions the active-force-length-curve of the Millard2012 model is substantially weaker than the Thelen2003 model, which could explain the loss of strength that you are seeing. You might find it helpful to check out the plots comparing the characteristic curves of the different models on confluence (http://simtk-confluence.stanford.edu:80 ... don+Curves).
If this is the case fear not: you can adjust the shape of the active-force-length-curve by changing its parameters using the C++ API. I think its also possible to use the property editor in the GUI to make this change as well. See the doxygen (https://simtk.org/api_docs/opensim/api_ ... ml#details) for the ActiveForceLengthCurve for details on what the properties mean and for details on the constructor you'd use to make your own ActiveForceLengthCurve.
Cheers,
Matt
Have you checked the fiber lengths and the value of the active-force-length multiplier? My guess is that the normalized fiber length is less than 0.6 or greater than 1.5. In these regions the active-force-length-curve of the Millard2012 model is substantially weaker than the Thelen2003 model, which could explain the loss of strength that you are seeing. You might find it helpful to check out the plots comparing the characteristic curves of the different models on confluence (http://simtk-confluence.stanford.edu:80 ... don+Curves).
If this is the case fear not: you can adjust the shape of the active-force-length-curve by changing its parameters using the C++ API. I think its also possible to use the property editor in the GUI to make this change as well. See the doxygen (https://simtk.org/api_docs/opensim/api_ ... ml#details) for the ActiveForceLengthCurve for details on what the properties mean and for details on the constructor you'd use to make your own ActiveForceLengthCurve.
Cheers,
Matt
-
Frédéric Lamon
- Posts: 27
- Joined: Wed May 06, 2015 1:01 pm
Re: moment vs angle plot for thelen and millard model
Hello Mr. Millard,
Thanks a lot for answering my questions. I was wondering if you had an idea as to which muscle models (equilibrium or acceleration) I should use for analyzing the tennis serve as it is what I am currently working on and I can't seem to figure out which of these two would be a better fit.
I also can't seem to figure out what is the active-force-length multiplier you suggested me to check and where to find it ?
I have also joined a picture as an example for my model with Thelen and your model to give you a better look at the differences. Because I was thinking on using these plots to figure out by how much I should strengthen my model in order to reach the maximal joint torque measured during a tennis serve.
Cheers,
Frederic
Thanks a lot for answering my questions. I was wondering if you had an idea as to which muscle models (equilibrium or acceleration) I should use for analyzing the tennis serve as it is what I am currently working on and I can't seem to figure out which of these two would be a better fit.
I also can't seem to figure out what is the active-force-length multiplier you suggested me to check and where to find it ?
I have also joined a picture as an example for my model with Thelen and your model to give you a better look at the differences. Because I was thinking on using these plots to figure out by how much I should strengthen my model in order to reach the maximal joint torque measured during a tennis serve.
Cheers,
Frederic
- Attachments
-
-
-
Matthew Millard
- Posts: 24
- Joined: Fri Aug 01, 2008 7:17 am
Re: moment vs angle plot for thelen and millard model
Hello Frederic,
As far as equilibrium vs. acceleration model is concerned in almost all cases the equilibrium model is a better choice: it will give you the same answers faster. If you're doing optimal control the acceleration model is a better choice because you can make fiber length, fiber velocity, and activation optimization variables since the dynamic equations leave these quantities un-coupled.
To understand what is causing the differences in joint torque between the Thelen2003Muscle and the Millard2012Equilibrium model we'd next need to have a look at plots of the normalized fiber length, and the normalized fiber velocity. You should be able to plot these quantities using the GUI just as you did for the joint torque plots. At the moment I can't verify this as I'm on a linux machine. If you can generate these plots then we can figure out what's causing the large difference in joint torques that you're seeing.
Matt
As far as equilibrium vs. acceleration model is concerned in almost all cases the equilibrium model is a better choice: it will give you the same answers faster. If you're doing optimal control the acceleration model is a better choice because you can make fiber length, fiber velocity, and activation optimization variables since the dynamic equations leave these quantities un-coupled.
To understand what is causing the differences in joint torque between the Thelen2003Muscle and the Millard2012Equilibrium model we'd next need to have a look at plots of the normalized fiber length, and the normalized fiber velocity. You should be able to plot these quantities using the GUI just as you did for the joint torque plots. At the moment I can't verify this as I'm on a linux machine. If you can generate these plots then we can figure out what's causing the large difference in joint torques that you're seeing.
Matt
-
Frédéric Lamon
- Posts: 27
- Joined: Wed May 06, 2015 1:01 pm
Re: moment vs angle plot for thelen and millard model
Hello Matt,
Once again, I would like to thank you for taking the time to look at my problem. I will proceed with the equilibrium model as suggested.
Concerning the plots, I was able to plot the normalized fiber length, but for the normalized fiber velocity, I tried to get the data using the analysis tool with muscle analysis, but the results are not coherent at all. I still attached a graph for thelen, but I can't seem to figure out why it is so large.
Cheers,
Frederic
Once again, I would like to thank you for taking the time to look at my problem. I will proceed with the equilibrium model as suggested.
Concerning the plots, I was able to plot the normalized fiber length, but for the normalized fiber velocity, I tried to get the data using the analysis tool with muscle analysis, but the results are not coherent at all. I still attached a graph for thelen, but I can't seem to figure out why it is so large.
Cheers,
Frederic
- Attachments
-
-
-