dorie issueshttps://ts-gitlab.iup.uni-heidelberg.de/dorie/dorie/-/issues2018-03-27T19:48:49+02:00https://ts-gitlab.iup.uni-heidelberg.de/dorie/dorie/-/issues/51Deploy source code to public repository2018-03-27T19:48:49+02:00Lukas Riedelmail@lukasriedel.comDeploy source code to public repository### Description
To make DORiE available to the public, we can push its source code to a public repository on the GitLab provided by the DUNE developers.
### Proposal
* Handle deployment as a stage in the CI/CD
* Deploy to Environments:
* Code itself
* Wiki, see #50
* push stable/release branches and master if
* they have been updated
* all tests pass
### To-Do
- [x] Create ~~public~~ _private_ repository: https://gitlab.dune-project.org/lukas.riedel/dorie
- [ ] Make repo public
### How to test the implementation?
Auto deployment works!### Description
To make DORiE available to the public, we can push its source code to a public repository on the GitLab provided by the DUNE developers.
### Proposal
* Handle deployment as a stage in the CI/CD
* Deploy to Environments:
* Code itself
* Wiki, see #50
* push stable/release branches and master if
* they have been updated
* all tests pass
### To-Do
- [x] Create ~~public~~ _private_ repository: https://gitlab.dune-project.org/lukas.riedel/dorie
- [ ] Make repo public
### How to test the implementation?
Auto deployment works!v1.0 Releasehttps://ts-gitlab.iup.uni-heidelberg.de/dorie/dorie/-/issues/43add options for altering DG formulation in local operator2018-03-15T15:06:00+01:00Lukas Riedelmail@lukasriedel.comadd options for altering DG formulation in local operatorThere are multiple options for different symmetry/penalty and upwinding combinations.
Upwinding:
- __fully weighted__ (`none`) (use $`K`$ for weighting)
- semi-upwinding (`semiUpwind`) (use $`K_0`$ for weighting and upwinding conductivity factor $`K_f^u`$)
- full updwinding (`fullUpwind`) (only _upwinding flux_ $`\mathbf{j}^u`$ upwinding conductivity $`K^u`$)
Symmetry & Penalty
- __SIP__ (`SIPG`) = Volume + Consistency + Symmetry + Penalty
- NIP (`NIPG`) = Volume + Consistency - Symmetry + Penalty
- OOB (`OOB`) (Oden, Babuska, Baumann) = Volume + Consistency - Symmetry
- IIP (`IIB`) (Incomplete Interior Penalty) = Volume + Consistency + Penalty
Harmonic weights:
- Weights on (`weights = true`): Use harmonic weighting of (saturated) conductivities at intersections
- Weight off (`weights = false`): No weightingThere are multiple options for different symmetry/penalty and upwinding combinations.
Upwinding:
- __fully weighted__ (`none`) (use $`K`$ for weighting)
- semi-upwinding (`semiUpwind`) (use $`K_0`$ for weighting and upwinding conductivity factor $`K_f^u`$)
- full updwinding (`fullUpwind`) (only _upwinding flux_ $`\mathbf{j}^u`$ upwinding conductivity $`K^u`$)
Symmetry & Penalty
- __SIP__ (`SIPG`) = Volume + Consistency + Symmetry + Penalty
- NIP (`NIPG`) = Volume + Consistency - Symmetry + Penalty
- OOB (`OOB`) (Oden, Babuska, Baumann) = Volume + Consistency - Symmetry
- IIP (`IIB`) (Incomplete Interior Penalty) = Volume + Consistency + Penalty
Harmonic weights:
- Weights on (`weights = true`): Use harmonic weighting of (saturated) conductivities at intersections
- Weight off (`weights = false`): No weightingv1.0 ReleaseLukas Riedelmail@lukasriedel.comLukas Riedelmail@lukasriedel.comhttps://ts-gitlab.iup.uni-heidelberg.de/dorie/dorie/-/issues/42improve Newton solver parameters2018-03-12T15:26:43+01:00Lukas Riedelmail@lukasriedel.comimprove Newton solver parametersThere were a couple of issues observed with the Newton solver:
1. `AbsoluteLimit` lower than 1e-8 is not observable in the (single precision) VTK output.
- ...but in mass conservation, see #39.
2. If `AbsoluteLimit` is too high, the solver will not perform an iteration for small time steps and weak dynamics.
3. We don't use `Reduction`.
Propositions:
- `AbsoluteLimit = 1e-10`
- `Reduction = 1e-4`: To save time, allow a lower defect if the reduction is sufficient
- `ForceIteration = true`: make sure that we don't skip dynamicsThere were a couple of issues observed with the Newton solver:
1. `AbsoluteLimit` lower than 1e-8 is not observable in the (single precision) VTK output.
- ...but in mass conservation, see #39.
2. If `AbsoluteLimit` is too high, the solver will not perform an iteration for small time steps and weak dynamics.
3. We don't use `Reduction`.
Propositions:
- `AbsoluteLimit = 1e-10`
- `Reduction = 1e-4`: To save time, allow a lower defect if the reduction is sufficient
- `ForceIteration = true`: make sure that we don't skip dynamicsv1.0 ReleaseLukas Riedelmail@lukasriedel.comLukas Riedelmail@lukasriedel.comhttps://ts-gitlab.iup.uni-heidelberg.de/dorie/dorie/-/issues/39write test for global mass conservation2018-10-07T18:13:08+02:00Lukas Riedelmail@lukasriedel.comwrite test for global mass conservationWe have to test if global water mass is conserved. To do so, we need a new local operator which computes $` V_w = \sum_{T} \int_T \theta_w `$. We can then compare the temporal change of this quantity to the fluxes into and out of the domain, $` d V_w / dt = \sum_{F} \int_F j_N `$.We have to test if global water mass is conserved. To do so, we need a new local operator which computes $` V_w = \sum_{T} \int_T \theta_w `$. We can then compare the temporal change of this quantity to the fluxes into and out of the domain, $` d V_w / dt = \sum_{F} \int_F j_N `$.v1.0 ReleaseLukas Riedelmail@lukasriedel.comLukas Riedelmail@lukasriedel.comhttps://ts-gitlab.iup.uni-heidelberg.de/dorie/dorie/-/issues/25revamp error estimator2018-03-12T16:43:42+01:00Lukas Riedelmail@lukasriedel.comrevamp error estimatorThe error estimation scheme will be adapted to the a posteriori flux error estimate by Di Pietro and Ern.
The flux error estimate across an intersection is given by the solution jump $`[u_h]`$ and the gradient jump. We extend the gradient jump to the jump in physical flux:
$`E_\text{flux} := C_{F, T}^{1/2} h_T^{1/2} \cdot || \overline{\omega}_F \left[ K_0 K_f \nabla_h u_h \right] \cdot \hat{\mathbf{n}}_F + \gamma \frac{\eta}{h_F} [ u_h ] ||_{L^2(\partial T)}`$
$` C_{F, T} := \left[ h_T |\partial T|_{d-1} |T|_d^{-1} \right] \left[ 2 d^{-1} + C_{P, T} \right] C_{P, T} `$
$` C_{P, T} := \pi^{-1} `$The error estimation scheme will be adapted to the a posteriori flux error estimate by Di Pietro and Ern.
The flux error estimate across an intersection is given by the solution jump $`[u_h]`$ and the gradient jump. We extend the gradient jump to the jump in physical flux:
$`E_\text{flux} := C_{F, T}^{1/2} h_T^{1/2} \cdot || \overline{\omega}_F \left[ K_0 K_f \nabla_h u_h \right] \cdot \hat{\mathbf{n}}_F + \gamma \frac{\eta}{h_F} [ u_h ] ||_{L^2(\partial T)}`$
$` C_{F, T} := \left[ h_T |\partial T|_{d-1} |T|_d^{-1} \right] \left[ 2 d^{-1} + C_{P, T} \right] C_{P, T} `$
$` C_{P, T} := \pi^{-1} `$v1.0 ReleaseLukas Riedelmail@lukasriedel.comLukas Riedelmail@lukasriedel.com