Why I am getting negative Total Energy when using preload springs (Hypermesh, Radioss)
Hello,
I am building a model in which I have plenty of preload springs that are deactivated at t=0. When a sensor is activated (generally distance between nodes becomes lower than a certain value) those springs start working.
The general behaviour in the animations seems to be logical. However, when I check the TH outputs, I see the Total Energy has decreased a lot, getting to greatly negative values. These values correspond to Internal Energy, in particular Spring Internal Energy corresponding to the springs that have been activated during the simulation.
In 0001.out file after simulation the %Error in energy is 0.00%, and the mass added is less than 5%.
I cannot make physical sense of this result, as I tend to think that the energy disipated in those springs should have been transferred to the attached elements via kinetic energy in rigid bodies, or deformation (ie internal energy) in shell elements. This is why I am concerned the dynamics I am seeing at the animations are not realistic, as those springs may not be working as intended.
Can anybody give a pyhisical explanation on this? Is my model still valid or am I missing the effects of the springs in the global model in any way? If so, how can I model my springs so that I remove this undesired effects?
Thank you in advance
Diego
Best Answer
-
Diego Caro said:
Hello Paul,
I attached a simplified model in which the same effect can be seen. I have one preloaded torsional spring and the Internal Energy quickly goes to negative values, but Contact Energy does not compensate and Total Energy is also negative at the end.
Stiffness is a known function for the spring.
Ah, ok, I thought you were using 'SPR_PRE' pre tension springs, but I see in your example you have KJOINT2 with non-zero moment for 0 rotation, this is a perfectly good way to define this, but not what I thought you were doing (stif0 is a value on the SPR_PRE card)
What you have discovered is 2 separate issues really, the fact that initial 'stored' energy in springs as you have them is not accounted for at t0 and the lack of any output for 'stonewall' energy in Radioss meaning that the rigid-rigid energy is 'lost' (in terms of 'plottable' energy at least), your energy output was very low frequency so I increased it so we can see better what happens in your example model:
Issue1: No accounting of stored energy in spring: If we look at the curves, KE and IE both start at 0 (showing the energy 'in' the spring is not accounted for), you can also see that as the Spring then releases energy, that stored energy being released is accounted for (but with a reference of 0 so it is negative), the lost energy in the spring is in fact (mostly) converted into the kinetic energy of the plates (as you expected) so long as they are moving the balance is maintained. until:
Issue 2: When the 2 rigid plates collide, the kinetic energy is lost very quickly (the sudden energy change at around 0.075) some is returned to the spring as internal energy reaching equilibrium with the angle/moment it has, but most becomes 'stonewall' which is not accounted for in TH plots in Radioss, so the overall balance jumps negative at that point.
The Stonewall is accounted for internally in Radioss however, If you look at energy error in the out file, you may see it is very good, at 0.0% because of this. I think your example model is fine, the energy balance may look odd in Time History, but it is explainable!
Regarding fixes/RFE, output of stonewall is already requested (not sure on if/when it will appear).
Accounting for spring energy is more difficult, since user could define a spring with e.g. constant moment regardless of Angle (i.e. a motor!) and area under the Spring curve would then be infinite and energy balance would then be even worse!
Hope this helps!
1
Altair EmployeeAnswers
-
Depending on your spring settings the energy balance may depend on Stif0 I think, can you share the model?
0 -
Paul Sharp_21301 said:
Depending on your spring settings the energy balance may depend on Stif0 I think, can you share the model?
Hello Paul,
I attached a simplified model in which the same effect can be seen. I have one preloaded torsional spring and the Internal Energy quickly goes to negative values, but Contact Energy does not compensate and Total Energy is also negative at the end.
Stiffness is a known function for the spring.
0 -
Diego Caro said:
Hello Paul,
I attached a simplified model in which the same effect can be seen. I have one preloaded torsional spring and the Internal Energy quickly goes to negative values, but Contact Energy does not compensate and Total Energy is also negative at the end.
Stiffness is a known function for the spring.
Ah, ok, I thought you were using 'SPR_PRE' pre tension springs, but I see in your example you have KJOINT2 with non-zero moment for 0 rotation, this is a perfectly good way to define this, but not what I thought you were doing (stif0 is a value on the SPR_PRE card)
What you have discovered is 2 separate issues really, the fact that initial 'stored' energy in springs as you have them is not accounted for at t0 and the lack of any output for 'stonewall' energy in Radioss meaning that the rigid-rigid energy is 'lost' (in terms of 'plottable' energy at least), your energy output was very low frequency so I increased it so we can see better what happens in your example model:
Issue1: No accounting of stored energy in spring: If we look at the curves, KE and IE both start at 0 (showing the energy 'in' the spring is not accounted for), you can also see that as the Spring then releases energy, that stored energy being released is accounted for (but with a reference of 0 so it is negative), the lost energy in the spring is in fact (mostly) converted into the kinetic energy of the plates (as you expected) so long as they are moving the balance is maintained. until:
Issue 2: When the 2 rigid plates collide, the kinetic energy is lost very quickly (the sudden energy change at around 0.075) some is returned to the spring as internal energy reaching equilibrium with the angle/moment it has, but most becomes 'stonewall' which is not accounted for in TH plots in Radioss, so the overall balance jumps negative at that point.
The Stonewall is accounted for internally in Radioss however, If you look at energy error in the out file, you may see it is very good, at 0.0% because of this. I think your example model is fine, the energy balance may look odd in Time History, but it is explainable!
Regarding fixes/RFE, output of stonewall is already requested (not sure on if/when it will appear).
Accounting for spring energy is more difficult, since user could define a spring with e.g. constant moment regardless of Angle (i.e. a motor!) and area under the Spring curve would then be infinite and energy balance would then be even worse!
Hope this helps!
1 -
Paul Sharp_21301 said:
Ah, ok, I thought you were using 'SPR_PRE' pre tension springs, but I see in your example you have KJOINT2 with non-zero moment for 0 rotation, this is a perfectly good way to define this, but not what I thought you were doing (stif0 is a value on the SPR_PRE card)
What you have discovered is 2 separate issues really, the fact that initial 'stored' energy in springs as you have them is not accounted for at t0 and the lack of any output for 'stonewall' energy in Radioss meaning that the rigid-rigid energy is 'lost' (in terms of 'plottable' energy at least), your energy output was very low frequency so I increased it so we can see better what happens in your example model:
Issue1: No accounting of stored energy in spring: If we look at the curves, KE and IE both start at 0 (showing the energy 'in' the spring is not accounted for), you can also see that as the Spring then releases energy, that stored energy being released is accounted for (but with a reference of 0 so it is negative), the lost energy in the spring is in fact (mostly) converted into the kinetic energy of the plates (as you expected) so long as they are moving the balance is maintained. until:
Issue 2: When the 2 rigid plates collide, the kinetic energy is lost very quickly (the sudden energy change at around 0.075) some is returned to the spring as internal energy reaching equilibrium with the angle/moment it has, but most becomes 'stonewall' which is not accounted for in TH plots in Radioss, so the overall balance jumps negative at that point.
The Stonewall is accounted for internally in Radioss however, If you look at energy error in the out file, you may see it is very good, at 0.0% because of this. I think your example model is fine, the energy balance may look odd in Time History, but it is explainable!
Regarding fixes/RFE, output of stonewall is already requested (not sure on if/when it will appear).
Accounting for spring energy is more difficult, since user could define a spring with e.g. constant moment regardless of Angle (i.e. a motor!) and area under the Spring curve would then be infinite and energy balance would then be even worse!
Hope this helps!
Thank you very very much Paul!
This has been super helpful!!
0