Learning Forum

[amurphy3485]

Hello,

I am running an Ansys transient simulation of flow around a bluff body to measure drag. I want to ensure the Cd has reached steady state, so I am trying to use convergence conditions.

Im using an area weighted average of pressure distribtution over the body to determine when steady state has been reached. My timestep is 0.0001s, with 100,000 timesteps to allow more than enough time for convergence to be reached. I want the solution to converge when this pressure residual reaches 0.001, ignoring the first 0.14 seconds (1400 timesteps) and taking an average value over 0.2 seconds of flow time (2000 iterations). Thus, I filled in the convergence condition as follows:

I left it running overnight, however the solution 'converged' at 0.398s (1398 timesteps). 

Examining the plot, the results residual definitely did not fall within 0.001 and did not even reach 2000 timesteps? Did I fill in the convergence conditions wrong

Thanks for you help.

[Rob]

The convergence criterion trigger the end of the time step, not the end of the solve. So, if the function is being checked, it's just terminating the time step. There isn't a good way to stop models with a transient feature that's no longer changing at present. Equally, if the monitors are expected to be single value why are you running transient?

[amurphy3485]

Is the solution convergence not supposed to check the overall solution convergence at the end of every timestep, and end the study if convergence has been reached? Also, I should have said pseudo-steady state. Its a transient study to capture the vortex shedding, so the lift forces fluctuate over time. Im trying to detemine when the flow devlopes and the oscillations become uniform.

[Rob]

No, it'll end the time step if convergence of that step is reached. 

[amurphy3485]

So is there no way of getting fluent to end the study once the flow become fully developed? It has to be done by watching the oscillations and visually determining that they are repeating?

Also, why would the convergence conditions have caused the solution to end at all? Should it not have have kept terminating the timesteps right up to 100,000?