Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.
Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
heaviside function application for pencil lead break test
Posted 2011年1月14日 GMT-5 11:34 Acoustics & Vibrations Version 4.2a 1 Reply
Please login with a confirmed email address before reporting spam
Hello Sir,
I am trying to get a simulation on 2D plate of aluminium, the pic attached, with some sensors on it to record the results.
I have to show an impact result on the plate on the top surface, with the help of an "Heaviside" function.
Can u please tell me how to place an equation using Heaviside and where to put, since it gives an error, may be an equation is wrong.
Tell me for creating an impact of force with (pencil lead break funda).. how to proceed..??
I am trying to get a simulation on 2D plate of aluminium, the pic attached, with some sensors on it to record the results.
I have to show an impact result on the plate on the top surface, with the help of an "Heaviside" function.
Can u please tell me how to place an equation using Heaviside and where to put, since it gives an error, may be an equation is wrong.
Tell me for creating an impact of force with (pencil lead break funda).. how to proceed..??
Attachments:
1 Reply Last Post 2011年1月14日 GMT-5 15:27
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hello
first of all a point load is strictly speaking "non physical" and leads to a stress singularity, which means that in the direct vicinity (3-10 local mesh radius) the results are very wrong at least for the true stress. But the global result might still be fully acceptable, and do not forget to do your verification/validation of the model by some other means (analytical) and the mesh sensitivity analysis in such cases. A way to improve things, is to apply the load onto a small finite surface/boundary (edge in 2D).
It is indeed a good idea to use a heaviside function to smoothen the dirac impact pulse, and to define a limited energy pulse, but a Gaussian pulse can also do, the main thing is to have a defined derivative during turn on and turn off, including helping the transient solver to sample the step with 3-10 points per rise AND per fall time.
You need perhaps a "strict" or "intermediate" stepping settings, rather tan the "automatic" that is tuned more for log type evolutions
The rest is traditional solver tuning, check your scalings that you do not have very different (1:10^6 or more differences) of your extreme depedent variable range. Check the initial values, the default "0" is not always the "best" starting piont, often we know far more, but are leasy, the result is often poor convergence and loss of our time
I always write my Heaviside cases as "functions" to be able to plot them to check that the shape is how I believe they are, and to help me define the time ranges for the solver settings
--
Good luck
Ivar
first of all a point load is strictly speaking "non physical" and leads to a stress singularity, which means that in the direct vicinity (3-10 local mesh radius) the results are very wrong at least for the true stress. But the global result might still be fully acceptable, and do not forget to do your verification/validation of the model by some other means (analytical) and the mesh sensitivity analysis in such cases. A way to improve things, is to apply the load onto a small finite surface/boundary (edge in 2D).
It is indeed a good idea to use a heaviside function to smoothen the dirac impact pulse, and to define a limited energy pulse, but a Gaussian pulse can also do, the main thing is to have a defined derivative during turn on and turn off, including helping the transient solver to sample the step with 3-10 points per rise AND per fall time.
You need perhaps a "strict" or "intermediate" stepping settings, rather tan the "automatic" that is tuned more for log type evolutions
The rest is traditional solver tuning, check your scalings that you do not have very different (1:10^6 or more differences) of your extreme depedent variable range. Check the initial values, the default "0" is not always the "best" starting piont, often we know far more, but are leasy, the result is often poor convergence and loss of our time
I always write my Heaviside cases as "functions" to be able to plot them to check that the shape is how I believe they are, and to help me define the time ranges for the solver settings
--
Good luck
Ivar