Discussion Forum

Posted: 1 decade ago 15 nov. 2011, 15:15 UTC−5

Hi use the step or ramp or pulse functions, and make them periodic (see the sub-tabs) or even use a sinus time oscillation velocity. One thing do observe that your functions must be smooth such that COMSOl might estimate their derivatives for the jacobian bilding, else the solver might well not converge for you. In any case start simple, add a layer by layer. anf finally input and output BCs are rather symmetrical so they mosty allow for bidirectional flow, depends a little on how you set up the BC specifications, keep them simple, in the worts casse you might need to write a few weak espressions yourself -- Good luck Ivar

Posted: 1 decade ago 15 nov. 2011, 17:22 UTC−5

Thanks Ivar, I tried using the ramp and step functions but I want to in fact have a BC at a point at time one and then have that point free with no BC at time two. Maybe I am using the step functions wrong but I am using them to set my BC's to a value (either 0 or 1). I was to find the option to set them to periodic. Is this in the function subtab? Alex

Posted: 1 decade ago 16 nov. 2011, 02:43 UTC−5

Hi you are right the periodicity seems to have disappeared in my 4.2a now, I do not believe I was dreaming. But on the other hand there are something new: "Waveforms" with sawtooth, sine, square, and triangles, these are periodic and smoothed -- Good luck Ivar

Posted: 1 decade ago 25 nov. 2011, 04:35 UTC−5

[QUOTE] I tried using the ramp and step functions but I want to in fact have a BC at a point at time one and then have that point free with no BC at time two. Maybe I am using the step functions wrong but I am using them to set my BC's to a value (either 0 or 1). [/QUOTE] Hello everybody, i think i have a similar problem. I try to displace the surface of a rod linear in time and then release it. What i expect is that the time dependent solution would give me a oscillating displacement. But instead of this i find some small damped oscillating displacements which vanish after extremly short time. The frequency of the displacement is dependent on the time-step of the solver. The boundary condition for the excitation is a prescribed displacement with the function (amplitude/delta_time)*t*(t

Posted: 1 decade ago 1 déc. 2011, 01:58 UTC−5

Hi if you use a prescribed displacement it will set the displacement at "0" that is on demand, you should rather use a force to displace it and then release the force, that is what you do in reality, no ? -- Good luck Ivar

Posted: 1 decade ago 6 déc. 2011, 06:40 UTC−5

Thank you, Ivar, for your response! If i use a boundary load instead of a prescribed displacement i can see the expected oscillation. Now I try simulate the necessary forces for the displacement in step1 with the prescribed displacement and with this results i set up a second calculation with the calculated forces as boundary load. can anyone give me a hint, where i can find information about this procedure?

Posted: 1 decade ago 7 déc. 2011, 01:30 UTC−5

Hi I believe, if I have understood you well, that you can add a general equation with a Force = AveeageDisp-DesiredDisp then apply the force to the boundary, calculate the average deformation of your sensor area as an aveop1(u) or whatever, define a parameter for DesiredDisp and let COMSOl adapt the force, you then plot the Force versus displacement. This is a nice way to apply moments. If you are not too much concerned of making the load surface "rigid" you might also use a "rigid Connector" and some of its subnodes This is used typically in solid, but I'm not sure how to introduce it correctly into your FSI problem. You might also drive the input pressure or velocity of your fluid (to get the desired deflection) -- Good luck Ivar