I'm solving the heat equation with time dependent boundary conditions numerically in a 2D system using the ADI scheme. For the purpose of this question, let's assume a constant heat conductivity and assume a 1D system, so ρcp∂T∂t=λ∂2T∂x2.
This works very well, but now I'm trying to introduce a second material. This one differs slightly in heat capacity and density but has a very different heat conductivity and is connected to the other material by a sharp interface, i.e. a stepwise change in λ.
How should this be treated numerically in the ADI scheme? I can think of different approaches:
- Treat the two materials as independent domains and connect them by a boundary condition calculating the heat flow in and out of the interface in terms of temperature on the other side of the interface in the last time step. Use a simple forward difference for that on both sides of the interface.
- Treat it as one domain and use a very fine discretization close the interface as compared to the homogeneous material. Use a scheme like λleftTi−Ti−1Δxi=λrightTj−Tj+1Δxj,where i and j are the points left and right of the interface, instead of the the standard ADI for those points.
- Drop the assumption of constant heat conductivity and use ρcp∂T∂t=∂λ∂x∂T∂x+λ∂2T∂x2.But in order do so one needs to approximate derivative of lambda at the step position, i.e. introduce an unknown characteristic width s of the sharp interface. I assume, that the (more or less arbitrary) choice of this width will significantly influence the system's behaviour.
Any advice?
No comments:
Post a Comment