The relaxation factor can speed up the calculation. In a passive air cooling simulation of the battery pack, I used the default parameter value (0.99) afforded by STAR-CCM+, but the temperature of battery pack is always rising and cannot be stabilized. But when I changed the parameter to 0.5, the temperature curve became stable.
As shown in Figure 1 and Figure 2, the relaxation factor is changed in 970s, the oscillation of the residual is also reduced. The temperature is also balanced.
According to the experience of your research, what range of relaxation factor can make the simulation get better results?
Thank you!
If the simulation converges, the relaxation factor should not affect the solution. As you say, it can affect how fast convergence happens. However, too high relaxation factors may destabilize the simulation, which will lead to wrong results. So "the relaxation factor does not affect results" only counts when the relaxation factor is low enough to lead to a converged solution.
If the simulation converges, the relaxation factor should not affect the solution. As you say, it can affect how fast convergence happens. However, too high relaxation factors may destabilize the simulation, which will lead to wrong results. So "the relaxation factor does not affect results" only counts when the relaxation factor is low enough to lead to a converged solution.
Before any conclusion, I suggest to run for longer time your simulation after the change done a t =960s
I think tha the relaxation factor does not affect results , but you can change it to converge your solution.It is between 0 to 1.
ends up affecting more easily the convergence of the results, so consequently affects yes.
Yes, the relaxation factors tend to simplify the complexity of solution. The complexity of solution tend to bring better results. So there might be deviation in results. In cfd fluent,
reducing momentum and dissipation rate will bring a change in results
1
Firstly check if the results get converged. An obtained result is not necessarily a converged result. A converged results has met the "Convergence Criteria",
2
If you get a converged result, you can be sure that the relaxation factor has not changed the accuracy.
3
If you have not get a converged result, the accuracy is not meaningful, since there is no "criterion"
Right relaxation factor speeds up the convergence of numerical solution and reduces the accumulation error in general. Relaxation factor choice depends on the type of differential equations( i mean for elliptic equation for example relaxation factor between 0.45-0.5). However, relaxation factor affects the degree of accuracy compared to experimental measurements.
Simulate with low relaxation factor first once the solution is stabilized, increase the urf and continue from stabilized solution for few more iterations.
- Badges
- Science topic