Learning Forum

[john.x.juveli]

All,

I'm running a Monte Carlo simulation emulating heating of a weld joint between two metals.

I would like to use adaptive time step methods, but it does appear that a variable of interest (e.g. the maximum tempearture of one of the metals) profile does change dependent upon how big/small the timestep is. It appears as though the maximum temperature reached over the sim is relatively less the larger the timestep is.

General question: Do I need to increase the # of target histories or how can I ensure the validity of my model and to ensure it is not dependent upon my timestep size? Am I perhaps missing some transients in between that are of importance and are not being captured in terms of aggregate energy here?

Any help is appreciated.

[Karthik Remella]

Hello The radiation estimated by the MC model depends on the random generation of photons in your simulation. If the number of target history is small and inadequate, you will tend to get a 'speckled' radiation field. Please check if this is true for your simulation. If this were true, then increasing the number of target history makes sense to get a smooth field.
Regarding your time step - did you calculate the time scales in your problem? You need to ensure that the time step you are selecting is smaller than the physical time scales in your problem. For example, if you were solving a conduction problem, your time step should be smaller than L^2/(\Alpha). '\Alpha' is the diffusivity of the material. This would be one way to make sure you are solving for the physics right.
Karthik

[john.x.juveli]

L^2 is the unit length of a mesh element?

[Karthik Remella]

L^2 is the length scale involved in your problem. Also, I just gave you an example of how you would go about estimating the timescale for a diffusion-dominated problem. You will have to similarly estimate for all the other physics you are attempting to solve.
Karthik