Learning Forum

[raju.chowdhury]

I am simulating compressible flow through a tiny annulus slot (1 mm width) between two pipes where gas is flowing at high pressure (20 bar). Initially gas is coming through a pipe of diameter 20mm and then gas is flowing through 1 mm annulus slot. Mesh of the computational domain is fine (all hexa mesh). Neither Simulation is not converging nor diverging. The velocity components residual fluctuating within a loop whereas other parameters reducing very slowly. I have run the simulation upto 60K iterations. After finishing 60K iterations, mass balance seems ok. Any idea, what could be the reason for not converging the simulation?n

[YasserSelima]

How large are the residuals? Can you show a plot?n

[raju.chowdhury]

Here it is. Residuals are set as default 1e-3.nn

[YasserSelima]

Not bad residuals. The reason you are getting this could be the nature of the problem itself. For example, if you have flow circulation or back flow, this is an unsteady problem by its nature. So, when you try to solve it in a steady simulation, you end up with oscillating residuals like what you have. However, the overall steady state solution you obtained could be a good average.nTry monitors, before and after the change in the area, and see if values oscillate or not.n

[raju.chowdhury]

Thanks for your reply.nI also tried with unsteady solver with time step size 1e-7. First 30 iterations converged well but after that not converged.n

[YasserSelima]

The unsteady solver will go up and down. Every time step, it goes up ... then converge. Is this what you mean?n

[raju.chowdhury]

No, I didn't mean that. For small time step size, the solution converged within 10 iterations (total iteration set 25 for each time step). After that, at each time step simulation finish all iterations (25) but no more converged. I monitored upto 1500 time steps. I expect that once the simulation will converge, the convergence will continue at every time step untill it solve all timesteps.n

[YasserSelima]

Not a rule. you might need to increase the number of iterations per time step.nAlso as you using small time step, used double precision if you don'tn