Atomization- floating point exception

- July 3, 2024 at 7:58 pm
  
  mayplain
  Subscriber
  
  I am currently simulating ultrasonic atomization by applying a dynamic mesh to the bottom wall. Although the simulation initially converges, I encounter a floating-point exception after approximately 200 time steps. The oscillation frequency is very high at 1.7 MHz, which may contribute to the error. However, I experience the same issue even when using a lower frequency. I aim to continue the simulation for a longer duration and would appreciate any suggestions for resolving the floating-point error. The mesh is structured and of very high quality. Additionally, I have reduced the time step to 1.84e-8, which should be sufficiently small. The dynamic was defined as sinusoidal oscillation with UDF. The explicit VOF with implicit body force was used to track the interface.
- July 4, 2024 at 9:47 am
  
  Rob
  Forum Moderator
  
  If you restart from an autosave does the model fail at the same point each time?
  - July 9, 2024 at 4:11 pm
    
    mayplain
    Subscriber
    
    I just tried to autosave and restart the model. It converged for 20-30 time steps after the first failing point. After that, it didn’t converge to my criteria but also didn’t diverge. After around 100 time steps, I received the floating point error again. I have attached the residual plot for your reference.
- July 9, 2024 at 4:56 am
  
  mayplain
  Subscriber
  
  Thanks for your response
- July 9, 2024 at 4:07 pm
  
  mayplain
  Subscriber
  
  I just tried to autosave and restart the model. It converged for 20-30 time steps after the first failing point. After that, it didn’t converge to my criteria but also didn’t diverge. After around 100 time steps, I received the floating point error again. I have attached the residual plot for your reference.
- July 12, 2024 at 9:30 am
  
  Rob
  Forum Moderator
  
  With VOF you may need a smaller time step, but you may also need more cell resolution as the liquid forms the ligands.
  - July 12, 2024 at 6:19 pm
    
    mayplain
    Subscriber
    
    I tried to use a very small time step and a very small mesh size, but still got the floating point error after some iterations. Do you have any other recommendations?
- July 15, 2024 at 10:39 am
  
  Rob
  Forum Moderator
  
  Please define small relative to the droplet/ligand sizes. If you look at the flow field just before divergence (that's why autosave is useful) what's happening?