[Muzz]

I am working on atmospheric boundary layer simulations and flow through an urban environment. I have been using realizable k-e with much success, utilizing boundary conditions and initial values etc. from best practice guidelines. However, whilst I can also obtain good results with standard k-e, when I make the switch to k-w SST (and k-w standard), I get a floating point exception (divide by zero) before the first iteration.

I believe I've correctly converted turbulence dissipation rate to specific dissipation rate for boundary/initial conditions. Is there a theoretical reason for this disparity in performance? What would a possible solution look like?

EDIT: I should add that this only occurs when using the all y+ wall treatment. When I switch to the high y+ wall treatment, the simulation runs as expected but underestimates the magnitude of turbulent kinetic energy throughout the environment.

[LuckyTran]

If you are using the same approach for the boundary conditions (intensity and length scale, or whatever) then it's likely your initial condition.

Originally Posted by Muzz

I believe I've correctly converted turbulence dissipation rate to specific dissipation rate for boundary/initial conditions. Is there a theoretical reason for this disparity in performance? What would a possible solution look like?

There isn't a disparity in performance. If it diverges in one iteration then either your settings are blatantly wrong or you didn't give a good enough initial guess.

You might want to try using field functions to convert the epsilon field into omega values when you switch models. If you are starting with a new case where you don't have a nice epsilon field, then you come up with better initial guesses.

[Muzz]

Originally Posted by LuckyTran

If you are using the same approach for the boundary conditions (intensity and length scale, or whatever) then it's likely your initial condition.



There isn't a disparity in performance. If it diverges in one iteration then either your settings are blatantly wrong or you didn't give a good enough initial guess.

You might want to try using field functions to convert the epsilon field into omega values when you switch models. If you are starting with a new case where you don't have a nice epsilon field, then you come up with better initial guesses.

Thanks for the response. The setup is identical when I change the turbulence model. All I'm changing is how I specify turbulence (K+Omega instead of K+Epsilon). I'm converting boundary/initial conditions from Epsilon to Omega to account for change in turbulence model.

What I don't get is how I can have good convergence with SKE and RKE but not have it work at all for KW-SST. What do you think the link is between the failure of the simulation and the wall treatment model? It works for High y+ but not All y+. I am using All y+ for SKE and RKE sims with no issue.

Thanks again for your contribution.

[Muzz]

After looking for answers on the steve portal, it recommends using -fpe in the command line when loading the .sim file, to generate more meaningful errors. The result is this:

SIGFPE: floating point exception
Command: Automation.Run
Recoverability: Non-recoverable
ServerStack: [SymInit: Symbol-SearchPath: '.', symOptions: 530, UserName: 'Desktop'
, OS-Version: 6.2.9200 () 0x300-0x1
, 00007FFEE2538BD4 (KwTurbModel): (filename not available): KwTurbLowReUstarEquation::KwTurbLowReUstarEquation
, 00007FFEE2519049 (KwTurbModel): (filename not available): KwAllYplusWallTreatment::setProperties
, 00007FFEE25179B1 (KwTurbModel): (filename not available): KwAllYplusWallTreatment::setProperties
, 00007FFEE250F102 (KwTurbModel): (filename not available): KwWallTreatmentModel::getWallTreatmentDomain
, 00007FFEFB3D9A26 (StarSolve): (filename not available): VerifiedFaceTerm::VerifiedFaceTerm
, 00007FFEFB3DB1FA (StarSolve): (filename not available): VerifiedFaceTerm::verifiedEvaluate
, 00007FFEFB3DEF3D (StarSolve): (filename not available): SelectiveFaceTerm::evaluate
, 00007FFEE2515E92 (KwTurbModel): (filename not available): KwAllYplusWallTreatment::accumulateTkeEquationWall Sources
, 00007FFEE272952B (KwTurbModel): (filename not available): KwTurbTkeBoundaryConditions::eval
, 00007FFEE2613349 (KwTurbModel): (filename not available): KwTurbModel::accumulateSources
, 00007FFEE2795442 (KwTurbModel): (filename not available): MenterKwTurbModel::accumulateSources
, 00007FFEE2718FC7 (KwTurbModel): (filename not available): KwTurbSlipVelocitySolver::getClassVersion
, 00007FFEE27186C5 (KwTurbModel): (filename not available): KwTurbSlipVelocitySolver::getClassVersion
, 00007FFEFB5385AB (StarSolve): (filename not available): RunnableSolver::doSolverIterationUpdate
, 00007FFEFB5371E1 (StarSolve): (filename not available): RunnableSolver::iterate
, 00007FFEFB54521F (StarSolve): (filename not available): SimulationIterator::startSimulation
, 00007FFEFB566413 (StarSolve): (filename not available): SimulationIterator::stopSimulation
, 00007FFF1867CA7F (StarNeo): (filename not available): Controller::executeCommand
, 00007FFF1867BB2C (StarNeo): (filename not available): Controller:: processCommands
, 00007FFF1A484779 (StarMachine): (filename not available): CommandController::scatterCommand
, 00007FFF1A483453 (StarMachine): (filename not available): CommandController:: processCommands
, 00007FFF1A5535EA (StarMachine): (filename not available): Machine::startServerHost
, 00007FFF1A53F618 (StarMachine): (filename not available): Machine::main
, ERROR: SymGetSymFromAddr64, GetLastError: 487 (Address: 00007FF6B4501EE0)
, 00007FF6B4501EE0 (star-ccm+): (filename not available): (function-name not available)
, ERROR: SymGetSymFromAddr64, GetLastError: 487 (Address: 00007FF6B450FEAF)
, 00007FF6B450FEAF (star-ccm+): (filename not available): (function-name not available)
, 00007FFF6D777BD4 (KERNEL32): (filename not available): BaseThreadInitThunk
, 00007FFF6DACCED1 (ntdll): (filename not available): RtlUserThreadStart

Does this mean anything?