Tuesday, July 13, 2021

How to... fix error Floating Point Exception -Overflow in Ansys CFX ?

 Many times, during the implementation of numerical modeling, after completing the boundary and initial conditions and starting the solver we have to do with a sudden stop of the analysis and the appearance of an overflow message.  This is the most common type of error that occurs when performing CFD analysis.  In today's post, I would like to present two paths to minimize the appearance of this error.

How to fix error floating point exception overflow ansys
Most often  error in CFX 


The first path will be the method most tested by me from the practical side.  It was effective in many cases where I encountered an overflow error.  The second path is the standard procedures to follow to get rid of this error.  In my opinion, two approaches should be familiarized with to have a full spectrum of knowledge about this error. It's written in such a general way by the solver, which makes it very difficult to diagnose directly.

To accurately describe and explain the first method, I will use an example of a modeling task.  So our problem will be cooling process of the plate at a pressure of 5 bar.  For this task, we will adopt the following assumptions: we will take into account the total energy model, we will define compressible air (ideal gas) as gas, we will define the SST turbulence model, we will take into account the radiation phenomenon by adopting the Surface to Surface model and that due to the symmetry of the model we will simplify the geometry to a periodic slice.

After defining the model phenomena, the time step was set at 1 s.  Then  solver was initiated. After couple of seconds simulation was stopped with an overflow error after seven iterations.  What next steps should be taken to locate an analysis failure?  Below I will present in order from the least labor-intensive to the most-labor-intensive activity.

1. Reduce the size of the time step.  The best reduction ratio is 1:10 scale.  So in our case we reduce the step from 1 s to 0.1 s. For transient analyzes, the definition of the correct time step size is much more important than the mesh quality.  The residual equations are more responsive to the size of the time step than to the change of the grid size.  In the case of static analyzes, we have such factors to change as the pseudo transient step or the time step of the solid domain.  If the next reduction to 0.01 s has no effect, proceed to the next step.

2. Change the gas model from compressible to incompressible.  This will cause our problem to reduce its nonlinearity.  Do not be afraid of this simplification if your problem concerns low pressures and speeds.  The differences in gas models in this case should not be large.  If this modification also does not help, go to the next one.

3. If the above two points do not work in the next step, turn off the radiation model. This should greatly simplify the analysis and the partial equations will be less susceptible to mesh quality. Radiation models significantly extend the simulation time. A bad mesh may lead to the formation of non-physical temperature values (values greater than the assumed temperature in the boundary and initial conditions).

4) The next stage is to reduce the value of the assumed boundary and initial conditions. If in the analysis you have assumed a pressure of over 2 bar or values close to the phenomenon of vacuum, change this value to atmospheric pressure. Also, when it comes to temperature values (above 1200 C), reduce them by at least half. The same should be done for the initial speed (if you are analyzing high speeds over 150 m / s).

5) If all of the above steps fail, perform the last step. When you have defined symmetry or periodicity, try to do calculations for the entire model without these boundary conditions. It can be quite a laborious step requiring the reconstruction of the model, but it can bring measurable benefits in the implementation of the analysis.

After completing the above steps, if you still have a problem with the error overflow: floating point exception, try the next methods posted on the CFD Online website. Below is a link to these materials. It is worth reading the entire post because the information contained therein is extremely valuable.

Ansys FAQ -- CFD-Wiki, the free CFD reference (cfd-online.com)








No comments:

Post a Comment