cfd in turbomachinery
 

Hy friends,

I try to simulate a centrifugal pump with fluent5 and mrf and mixing-plane. fluent averages for this purpose in circumferential-direction by "area-averaging". I do not know if this is ok. isn't it better to do this by mass-averaging or in an other way, so that all basic equations are "correct" over the mixing-plane or is this (because of theoretical reasons) not possible ? the different averaging-scemes give for the velocities really different values, and some differ about 100%.

thanks a lot tom

Re: cfd in turbomachinery
 

(1). Regardless of the code you use, the real problem is 3-D transient. (2). Using mixing-plane averaging is like putting a layer of screen there to get uniform profiles after the averaging. (3). The down-stream profile will be uniform, but it also affects the up-stream flow field. (4). The mixing-plane approach is not the ultimate solution, it is simply a man-made condition to link two separate flow fields in one computation to avoid the real 3-D transient computations. It is artificial conditions. For this reason, it is up to you to do the averaging. (5). The other way to do is to compute the component flow fields separately with extended up-stream and downstream boundary conditions. Then pick a station in the overlapping region and perform the averaging. This averaged condition can then be used as the inlet condition for the next flow field region.

Re: cfd in turbomachinery
 

Large differences between area and mass averaging are often caused by areas of recirculation, with some mass flow going in the 'negative' direction.

Re: cfd in turbomachinery
 

(1). It is always a good idea to check the 3-D flow field first. (2). Flow separation can be a big problem in any simulation, especially when trying to figure out the proper computational domain. (3). A confined separation bubble is not a problem. An open separation extending across the boundary can be difficult to handle.

Re: cfd in turbomachinery
 

Hi, My researches have been ever focus on interface techniques for multistage axial turbomachinery flow. As I know, It must be calibrated with experiment data before a multistage code could be used into design. As to which average could be used in interface treatment, less suggetion can be provided. For higher loaded and more farther departure from design point and more blade rows are included, mixing interface with any average techniques could be difficult for reasonable solution. Time-average model is a better selection, also Adamczyk's passage average equations are best directions. Another suggestion is that you'd better to do a time-accurate solution first, it will be a sound basis.

best wishes Ji

Re: cfd in turbomachinery
 

Some people seem to favor mass averaging for things like total pressure and temperature. Others like a "stream-thrust" average (where a non-uniform profile is mixed out in a 1D sense). As you know, these averages aren't good if there's any recirculation accompanying backflow at the interface. It seems to us that the area average seems to be the safest approach, though it may not be optimal in many cases. Providing a variety of averaging methods doesn't seem to be a bad idea so that people can pick and choose. Is that what you want ? I can't promise to put it in our product right away. But I wonder if doing that meets your needs.

Re: cfd in turbomachinery
 

Hy,

the possibility to choose the averaging-sceme is ok.

I read some time ago (sorry, I do not know where, sorry) an article, where a different approach is used. They integrate somehow over the mixing-plane by using a finite-volume-sceme. The result is, that continuity and (!!) the Navier-Stokes-Equations (incompressible) are "true" over the mixing-plane.

tom