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.
2D Heat Transfer with Sinusoidal Boundary Condition. Need Guidance
Posted Aug 25, 2015, 2:54 p.m. EDT General, Heat Transfer & Phase Change, Materials 1 Reply
Please login with a confirmed email address before reporting spam
Hi All,
I am building a 2D model with a heat source on the sides that provides fluctuating heat to the system by a sinusoidal function P=0.5[W]*sin(60[Hz]*t)). I am computing a time dependent study from 0 to 1000 seconds with a time step of 0.1. My main interest is the temperature curves at the center of the model, so I have in place point evaluations where I need them.
I've been running several time studies, varying the bounds and step, trying to determine how the accuracy of the simulation changes. I'm interested to see what happens at equilibrium so I want the simulation to show so I've been asking the simulation to run in steps of 10 or 100 for the first 10000 seconds and then pulling a 50 to 100 seconds of very precise calculations.
My problem so far is that the plots have been very rough when using a sinusoidal heat source. It either looks like a sawtooth plot or is extremely chaotic. But when I remove the sinusoidal in favor of a linear or exponential function, the curve looks like I would expect.
Does anyone see an issue with my function that I can't or knows the issue I'm dealing with and can provide assistance.
Much appreciated
- Charlie
I am building a 2D model with a heat source on the sides that provides fluctuating heat to the system by a sinusoidal function P=0.5[W]*sin(60[Hz]*t)). I am computing a time dependent study from 0 to 1000 seconds with a time step of 0.1. My main interest is the temperature curves at the center of the model, so I have in place point evaluations where I need them.
I've been running several time studies, varying the bounds and step, trying to determine how the accuracy of the simulation changes. I'm interested to see what happens at equilibrium so I want the simulation to show so I've been asking the simulation to run in steps of 10 or 100 for the first 10000 seconds and then pulling a 50 to 100 seconds of very precise calculations.
My problem so far is that the plots have been very rough when using a sinusoidal heat source. It either looks like a sawtooth plot or is extremely chaotic. But when I remove the sinusoidal in favor of a linear or exponential function, the curve looks like I would expect.
Does anyone see an issue with my function that I can't or knows the issue I'm dealing with and can provide assistance.
Much appreciated
- Charlie
1 Reply Last Post Aug 26, 2015, 4:07 a.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Hi
there are two points I would stress:
1) numerical solving issues: when you have an oscillatory boundary load, particularly in diffusion problems such as HT you should set the solver time stepping to "intermediate" to ensure it takes enough steps to catch your oscillations, then if you have a 60 Hz oscillation and you want to resolve this frequency, you need at least 5 points per period, hence a time stepping of Delta_t <= 1/(5*60[Hz]) !
2) HT physics, check your heat diffusivity (alpha) for your material you will notice that a 60 Hz oscillation of the boundary temperature will only penetrate a few, to some hundred microns, or at most a mm into the material, thereafter there are no "oscillatory response" you get only a smooth heat profile
So probably you could make your simulation quicker by applying a RMS power as the input flux and do a stationary solving
Heat transport is a diffusion equation, it behaves very differently from a classical wave equations
--
Good luck
Ivar
there are two points I would stress:
1) numerical solving issues: when you have an oscillatory boundary load, particularly in diffusion problems such as HT you should set the solver time stepping to "intermediate" to ensure it takes enough steps to catch your oscillations, then if you have a 60 Hz oscillation and you want to resolve this frequency, you need at least 5 points per period, hence a time stepping of Delta_t <= 1/(5*60[Hz]) !
2) HT physics, check your heat diffusivity (alpha) for your material you will notice that a 60 Hz oscillation of the boundary temperature will only penetrate a few, to some hundred microns, or at most a mm into the material, thereafter there are no "oscillatory response" you get only a smooth heat profile
So probably you could make your simulation quicker by applying a RMS power as the input flux and do a stationary solving
Heat transport is a diffusion equation, it behaves very differently from a classical wave equations
--
Good luck
Ivar