What are the 3 types of rocks involved in the rock cycle?

Thank you Muddasar Anwar , you have been vert helpful. I will definitely try this approach.

Hi.

Although it is not explicitly said in the question, I assume you are using RANS approach with mixing planes. And so, you are worried about the proximity of the inlet and/or outlet boundary condition (BC) to your mixing plane, i.e you are looking for the minimum distance to simulate to minimise 1/ the number of cells (CPU cost) and 2/ the influence of the BC to your result? I'm I correct?

The answer really depends of the physics you are taking into account and is mostly case dependant. In that senses, Muddasar answer is the right way to go. Potential effect can already be observed with a coarse mesh. Pumps are not within my expertise but I think the correct referential length should be the blade chord and not the pump diameter if your concern is the potential effect, since it is driven by the blade. From my experience, distance from the BC to the blade must be at least half a blade chord.

I may recommend the paper of Benichou et. al (Application of an Analytical Method to Locate a Mixing Plane in a Supersonic Compressor GT2014-25103). Although it is in the context of supersonic centrifugal compressor, some ideas could be used in your case.

Best.

Dear Jerome Dombard ,

Thanks for your answer. Yes, I am using RANS approach with mixing planes (more precisely, sliding mesh as I am running unsteady simulations as well).

You are correct in your statements, as I am interested on having BC specified properly with the minimum number of cells.

Since I posted this question I have run a series of steady simulations and I noticed that the velocity streamlines at the inlet are perfectly straights after 1D (I mean, they are parallel to the centerline of the pipe and orthogonal to the inlet faces). On the other hand, at the outlet, the streamlines are swirled because of the impeller rotation and this occurs also when I set the pipe length to 10D, but I don't experience backflow after 5D.

An inlet pipe which length is 1D seems to me a bit short, despite it would save me a lot of cells (I am currently at 5kk with 10D both pipes).

However, I think I cannot prevent on having swirl at the exit because of the impeller rotation, as I would need a pipe of infinite length. So I should only focus on not having backflow at the outlet. Am I correct?

If I am correct, I think I should go for 2D length for the inlet pipe and 5D for the outlet pipe, or maybe you have another set of lengths that you would recommend?

(Let's consider D as pipes diameter or chord length, as they are almost the same)

My advices may not apply since my expertise only rely on unsteady high fidelity simulations (LES approach) and non reflective boundary conditions (with a relax). The common rule is to avoid recirculation zones crossing boundaries and to be relatively independent of the geometry.

It's not only sufficient to have parallel streamlines at the boundary since you could still have a mean 2D profile of mass flow rate due to the potential effect of your blade. Look at the fields or transverse plane (like the inlet), especially pressure and mass flow rate (the two are related).

Is your mesh structured or unstructured? If unstructured, you could have a longer inlet and outlet and set your mesh to be coarser at the extremities so the numerical dissipation of your convective scheme may dissipate the smallest scales. At least in LES.. It may be not relevant in RANS since you do not solve turbulent eddies.

I think I have understood now. Thanks Jerome.

However I have a pump with asymmetric impeller (single vane) with double curvature volute (wastewater pump). I am working with unstructured mesh at the moment, as having a structured grid with this geometry is very challenging, but I am working on it, as I want to get a y+=1 to evaluate the possibility of LES at the very end of the project. I might need your expertise for that moment. But at the moment I am solving RANS.

Before LES, I would suggest to go as quickly as possible toward URANS + mixing planes.

Concerning your comment about y+=1 and LES, note that wall modelled LES (i.e with a standard law-of-the-wall) provide fairly accurate results on real configurations (i.e more complex than a flat plate) and can be considered as a first realistic and affordable approach vs wall-resolved LES. See CERFACS papers with the AVBP code on turbomachine configuration on that matter.

Don't hesitate to contact me.

Best,

Hi Muddasar Anwar ,

Thanks again for getting back to me. My comment about the infinite extension of the pipe was from a mathematical point of view, but I got the point on the inconvenience of having a pipe too long, as my study is more focused on the power performance (head, power, efficiency). However, my model should be coherent with the physics I am investigating. Regarding this, I am more interested on what is happening in the rotating regions (turbulence mainly), so I am not worried about what happens too far from it. I just have to be sure that I don't have backflow and my BC are not influenced by the rotating region. According to this, I should be fine with 5D upstream and downstream if I have well understood.

Answering to Jerome Dombard , I am already on uRANS and sliding mesh for almost two years now. On my simulations, I set the pipe length to 10D, but from a paper that I was reading the author used 4D, so I came up with this trivial question because I wasn't considering to shorten the pipe before.

However, LES are at an interstellar distance for me at the moment, as the geometry is very complicated and my mesh skills are still very poor to generate a grid that allows me to get a y+ approaching one. My PI advised me to look at Chimera and levelset mesh (I don't even know if I wrote that correctly), but I have started googling today. Do you have any recommendation about guides? I am also wondering if the integrated meshing software is fine or I should switch to ICEM or an external software for grid generation.