The easiest and most straight forward solution that works almost everytime is to enhance your mesh quality, make it as structured as possible and avoid overconstraining near the BC. Investing some time on your meshing certainly improves your solution accuracy and time by multiple folds.

Good Luck!

Hi

Computational cost depends on step size, element size and the number of iterations.

First, you have to start with the coarse meshing. If your solution not converged, go for a medium then fine meshing.

Secondly, you have to choose the time step to be one second. After that, you can reduce the step size.

Thirdly, you can play with the number of iterations to get a converged solution.

So basically you have to perform a grid independence test for your numerical model. You can change the above parameters to get a better-converged solution. However, more convergence might take more computational time, depending on your model.

In a specific problem, I found that my model which was dealing with the heat transfer of a ground heat exchanger, different domains had different meshing. The pipe required a fine meshing, whereas the ground required a coarse meshing. In that way, I was able to reduce computation time.

Hope my answer will helpful for you. If you need any further information/help, don't hesitate to write me.😊👍🏻

Dear Nima Khaledy

Keeping the orthogonality of the grids constant, make their sizes a bit larger.

I agree with Subhabrata answer

Hi Nima,

Short answer is a very good initial condition. The better it is the faster you obtain your final solution. Furthermore, all hexa elements are better than any other types of elements (just solely based on my personal experience). Of course there are several other factors involve in such simulations. Do not forget to monitor your mass imbalance since convergence based on residuals are not enough.

Regards,

Shervin

Since the accumulation of grid size over the wall and boundary layers is relatively higher than every other region, I would offer increase Y+ as long as results are in the range.

There are various ways to decrease the computation cost in CFD problem as below:

- Use a coarse meshing in unimportant sections of domain,

- Apply a more powerful processor,

- Utilize all processors of your system.

Hi Nima Khaledy. I don't think it is easy to simplify a multiphase water-air flow problem to one phase, much less in Ansys Fluent. You must take into account that that the VOF model is generally used, which in turn requires a refined mesh mainly near the free surface (which is possible we can hardly intuit where such a free surface is located). If you add a turbulence model to the above (for example k-e RNG, k-w SST), the computational cost increases greatly. I must also tell you that convergence is difficult to achieve in several cases.

I don't know what type of flow problem (open channel, culvert, flow over a weir, stepped channel, flow through a gate, etc.) you are solving, but I can think of the following general recommendations:

1) In ANSYS Fluid:

1.1) Run the simulation with double precision and several processors in parallel (depends on your laptop or similar).

1.2) If the flow can be analyzed as Steady I recommend: model it as Steady; VOF with Implicit formulation; Turbulence with k-e RNG (ensuring that y +> 50, so that you do not resolve the viscous sublayer, otherwise you should to use improved Scalable or Enhanced Wall functions or even the k-w SST model with y +