Hi,
I am simulating an axial turbine. I am simulating an axial turbine by using Fluent. The convergence of solution is monitored by checking the residuals of the numerically solved governing equations. Moreover, in order to judge the convergence, the behaviour of other quantities, such as the total pressure at the inlet and outlet boundaries, and torque coefficient generated by the rotor, are also monitored. Here, the default convergence criterion of each residual is reduced in order to allow the monitored quantities to stagnate at consistent values.
Under relaxation factors are default. Solver is ansys Fluent. This turbine is named Wells turbine. Solution is steady. The computation was carried out with changing the rotational speed of the rotor from a high speed to a low speed under a constant axial velocity.
Residuals are as follows:
http://upsara.com/images/ak6c_residuals3.jpg
http://upsara.com/images/nlfk_residuals4.jpg
http://upsara.com/images/qif_residuals5.jpg
And also, mass flow rate, torque (in direction of Y) and inlet and outlet total pressure are as follows(in order to judge on convergence):
Torque (in direction of Y):
http://upsara.com/images/8q40_y_momment3.jpg
http://upsara.com/images/zf53_y_momment4.jpg
Inlet and outlet total pressure:
http://upsara.com/images/89os_inlet_and_outlet_pressure3.jpg
http://upsara.com/images/uep0_inlet_and_outlet_pressure4.jpg
Mass flow rate in inlet and outlet:
http://upsara.com/images/5ep1_mass_flow_rate3.jpg
Now, I have some question about Residuals:
As you see in first figure, residuals decrease until about iteration of 1500 and then after iteration of 1500, residuals increase. I want to know that this event is logical or not. I already thought that residuals should decrease decidedly because of convergence. But in this simulation, residuals increases and decreases!! What is reason for this event?
According above figures, Is solution converged?
I am grateful that guide me about these cases.
Thanks.
So there all still problems with unstability at local Reynolds numbers >2. That I learned when I worked with FDM -methods. The problem could be solved with so-called upwind methods. But unfortunately that caused "false" diffusion or increased viscosity. A finer mesh could be a solution. A method was developed by D.F.Roscue that was able to use different degree of up-winding' I.e. at local Re = 0 it lead to central differencing and at Re= very high to 100 % upwind differencing.
- Badges
- Science topic