- Thank you received: 6
DG in time with a slab solver
- Guosheng Fu
- Topic Author
- Offline
- Elite Member
Less
More
3 years 7 months ago #3710
by Guosheng Fu
DG in time with a slab solver was created by Guosheng Fu
Hello,
I am exploring DG time solvers. In particular, I have a simple diffusion equation
u_t - u_xx=0.
I want to use CG to discretize in space and upwinding DG in time. I think the TensorProductFespace is designed for such problem, but I failed to produce a working code. (the tutorial examples in tensorproducts are all explicit DG...)
Mathematically, we shall be able to solve the linear systems in time slabs, I don't know whether this would be possible to realize in the current TensorProductFEspace. Can you help me with this? Many thanks!
Best,
Guosheng
I am exploring DG time solvers. In particular, I have a simple diffusion equation
u_t - u_xx=0.
I want to use CG to discretize in space and upwinding DG in time. I think the TensorProductFespace is designed for such problem, but I failed to produce a working code. (the tutorial examples in tensorproducts are all explicit DG...)
Mathematically, we shall be able to solve the linear systems in time slabs, I don't know whether this would be possible to realize in the current TensorProductFEspace. Can you help me with this? Many thanks!
Best,
Guosheng
3 years 7 months ago - 3 years 7 months ago #3711
by schruste
Replied by schruste on topic DG in time with a slab solver
As far as I know the TensorProductFESpace is not directly made for this.
In ngsxfem we have TensorProduct Space-Time spaces (also fitted), see e.g. here:
github.com/ngsxfem/ngsxfem/blob/release/...pacetimeDG_fitted.py
You can try it out without installation through binder here:
mybinder.org/v2/gh/ngsxfem/ngsxfem-jupyt...acetime_fitted.ipynb
Best,
Christoph
In ngsxfem we have TensorProduct Space-Time spaces (also fitted), see e.g. here:
github.com/ngsxfem/ngsxfem/blob/release/...pacetimeDG_fitted.py
You can try it out without installation through binder here:
mybinder.org/v2/gh/ngsxfem/ngsxfem-jupyt...acetime_fitted.ipynb
Best,
Christoph
Last edit: 3 years 7 months ago by schruste.
- Guosheng Fu
- Topic Author
- Offline
- Elite Member
Less
More
- Thank you received: 6
3 years 7 months ago #3712
by Guosheng Fu
Replied by Guosheng Fu on topic DG in time with a slab solver
Thanks for the examples.
- Guosheng Fu
- Topic Author
- Offline
- Elite Member
Less
More
- Thank you received: 6
3 years 7 months ago #3715
by Guosheng Fu
Replied by Guosheng Fu on topic DG in time with a slab solver
Hi Christoph,
Your spacetime framework is exactly what I am looking for. I have an additional question: I want to test ideas in 1D first as my model is a bit complicated. But your framework seems only support dimension d>=2. Is there a easy way around?
I am thinking of directing work on a tensor-product quad slab domain, but some of the finite elements seems missing: I need two nonstandard fe spaces: Pk-in-x direction and P-0 in y-direction (time), also Pk-in-x direction and P-k-1 in y-direction.
Are these spaces already available in ngsolve core?
Best,
Guosheng
Your spacetime framework is exactly what I am looking for. I have an additional question: I want to test ideas in 1D first as my model is a bit complicated. But your framework seems only support dimension d>=2. Is there a easy way around?
I am thinking of directing work on a tensor-product quad slab domain, but some of the finite elements seems missing: I need two nonstandard fe spaces: Pk-in-x direction and P-0 in y-direction (time), also Pk-in-x direction and P-k-1 in y-direction.
Are these spaces already available in ngsolve core?
Best,
Guosheng
3 years 7 months ago #3717
by schruste
Replied by schruste on topic DG in time with a slab solver
Hi Guosheng,
I just added 1D, at least for the scope of the examples that I've send you. You just need to update ngsxfem (master-branch). With that a script as attached works: .
Best,
Christoph
I just added 1D, at least for the scope of the examples that I've send you. You just need to update ngsxfem (master-branch). With that a script as attached works: .
Best,
Christoph
Attachments:
Time to create page: 0.109 seconds