[maebr]

Hi there,

We are designing the special centrifugal compressor for a specific application. So, we should do it A to Z.
Now, we have a centrifugal compressor, following with a vaneless diffuser. We chose the total inlet pressure and mass flow rate as the known boundary conditions for the inlet and outlet, respectively. Also, the total energy option has been selected for the heat transfer of both the impeller and the diffuser. It is worth mentioning that the interface between the impeller and diffuser is chosen as the frozen rotor. The problem is that the total temperature drops around 10 degree from diffuser inlet to outlet in steady state flow, while it should remain constant since no work in done in diffuser. We checked several things including: refining the mesh especially in interfaces, changing the outlet boundary condition to static pressure, changing the interface to stage, etc.
I appreciate if someone can give me some ideas about the possible reasons.

[Jiricbeng]

I do not think it should be something unexpected. Think of a following situation: a centrifugal compressor with total pressure at diffuser inlet - 4bars and total pressure at diffuser outlet 3.85 bars has a total pressure loss across the diffuser 0.15bar. The diffuser total inlet temperature is 450 K. If you calculate total diffuser outlet temperature (suppposing adiabatic expansion), you obtain 445 K. The diffuser outlet temperature is a function of aerodynamic desigh of the diffuser. The higher total pressure loss, the lower gas total temperature at the diffuser outlet.

[Opaque]

Agreed.

It does not have to be a diffuser, for example, the classic Fanno problem, i.e. compressible flow within an adiabatic pipe with friction. The total temperature does not remain constant, it will only if the friction factor = 0, i.e. no losses.

[maebr]

Thank you for your answers. But we are talking about the total temperature which should be constant in the diffuser since no work is done in it. (h02=h01 in difuser), and sth you said is correct but for the static pressure and temperature. As all we know, we are using diffuser after the compressor to increase the static pressure by decreasing the kinetic energy (and so velocity) while the total pressure decreases. In adiabatic situation, while there is not any work done in diffuser, why the total temperature changes?

[Opaque]

ok. Got it.

What are your settings for the heat transfer model, besides using Total Energy?

[Jiricbeng]

Originally Posted by maebr

As all we know, we are using diffuser after the compressor to increase the static pressure by decreasing the kinetic energy (and so velocity) while the total pressure decreases.

Again, I think that since the total pressure is decreased, the total temperature must also decrease in adiabatic process, in accordance to the equation as shown below, where all the variables are "total":
(T2/T1) = (p2/p1)^[(K-1)/K].



Sure, there is no work done - the equation above does not consider any work neither. It describes the adiabatic process only.

Because there is the total pressure loss, then T2<T1. If there was not total pressure loss (no friction, no little vortex, no flow bending etc.), then T2 = T1.

[maebr]

Originally Posted by Opaque

ok. Got it.

What are your settings for the heat transfer model, besides using Total Energy?

The walls are adiabatic. As you said the total energy is selected as the heat transfer option. Furthermore, we chose Shear stress Transport option for Turbulence model with the automatic wall function. It is worth mentioning that during running there were a notice as follow:
| ****** Notice ****** |
| A wall has been placed at portion(s) of an OUTLET |
| boundary condition (at 2.3% of the faces, 2.1% of the area) |
| to prevent fluid from flowing into the domain. |
| The boundary condition name is: Outlet. |
| The fluid name is: Fluid 1. |
| If this situation persists, consider switching |
| to an Opening type boundary condition instead.

which the outlet is the diffuser outlet where we have the problem with the presented results (amount of total temperature which should remain constant from diffuser inlet to its outlet.)

[maebr]

Originally Posted by Jiricbeng

Again, I think that since the total pressure is decreased, the total temperature must also decrease in adiabatic process, in accordance to the equation as shown below, where all the variables are "total":
(T2/T1) = (p2/p1)^[(K-1)/K].



Sure, there is no work done - the equation above does not consider any work neither. It describes the adiabatic process only.

Because there is the total pressure loss, then T2<T1. If there was not total pressure loss (no friction, no little vortex, no flow bending etc.), then T2 = T1.

Since there in not any work done in diffuser it means that:
h01=h02 where 1 and 2 are the diffuser inlet and outlet, respectively. So:
h02-h01=cp(T02-T01)=0
T02=T01 in diffuser
and the equation you mentioned is used for static pressure and temperature. since,
T01=T1+C1^2/(2*cp)=T2+C2^2/(2*cp)
which means that in diffuser that the velocity decreases, the static temperature increases and so the static pressure, while the total pressure decreases.

[maebr]

Originally Posted by Jiricbeng

Again, I think that since the total pressure is decreased, the total temperature must also decrease in adiabatic process, in accordance to the equation as shown below, where all the variables are "total":
(T2/T1) = (p2/p1)^[(K-1)/K].



Sure, there is no work done - the equation above does not consider any work neither. It describes the adiabatic process only.

Because there is the total pressure loss, then T2<T1. If there was not total pressure loss (no friction, no little vortex, no flow bending etc.), then T2 = T1.

And you are right that in the real case since the walls are not completely adiabatic and there are some losses, the total temperature decreases a little (in the most cases I checked even less than 1 degree) but in our case, the total temperature drop is more than 16K from diffiser inlet to outlet while the total pressure drop is around 1.4 bar.

[Opaque]

Ok. More explicit in my question, what is your setting for viscous work?

[Jiricbeng]

Originally Posted by maebr

Since there in not any work done in diffuser it means that:
h01=h02 where 1 and 2 are the diffuser inlet and outlet, respectively. So:
h02-h01=cp(T02-T01)=0
T02=T01 in diffuser
and the equation you mentioned is used for static pressure and temperature.

I guess h01 or h02 is enthalpy?
If yes, I do not think it is a correct assumption. Since in the diffuser no heat change is expected, the flow should be adiabatic. Isoenthalpic process occurs during throttling of a gas for example. Hence I am a bit suspicious about the equation h01 = h02 in case of diffuser. But maybe I am wrong.


As for the equation I mentioned, I ve seen it being used even for "total" variables.

[maebr]

Originally Posted by Opaque

Ok. More explicit in my question, what is your setting for viscous work?

In heat transfer, as I said I chose the total energy option and I forgot to mention that I selected the Inc.Viscous work term option. The turbulence option is SST with automatic wall function. The other options are not selected. All the walls selected to be smooth and adiabatic with no slip wall option of mass and momentum.
The problem is that the total enthalpy in the diffuser is not constant and changed from inlet to outlet. If there is anything else you need to know, please feel free to ask.
Furthermore, it seems that the notice is as a result of the short diffuser and since there is not any volute yet. So, I add a long duct after that to solve this problem which the notice disappeared. But, still we have the total temperature and total enthalpy drop in the diffuser. It seems that it is assumed that the total temperature at outlet should reach to the amount close to inlet temperature.

[maebr]

Originally Posted by Jiricbeng

I guess h01 or h02 is enthalpy?
If yes, I do not think it is a correct assumption. Since in the diffuser no heat change is expected, the flow should be adiabatic. Isoenthalpic process occurs during throttling of a gas for example. Hence I am a bit suspicious about the equation h01 = h02 in case of diffuser. But maybe I am wrong.


As for the equation I mentioned, I ve seen it being used even for "total" variables.

Yes, sorry I forgot to mention that the h01 and h02 are the total enthalpy at diffuser inlet and outlet, respectively. Yes, as you said since there is not any work done in the diffuser and no heat exchange, so the total enthalpy should remain constant in the diffuser. I can mention it here as a reference: page 138 of following reference book

Sayers, Anthony Terence. Hydraulic and compressible flow turbomachines. No. BOOK. McGraw-Hill, 1990.

As it can be found the total enthalpy is constant in a diffuser (Here the diffuser is following a centrifugal compressor's impeller). For the air assuming as an ideal gas then it can be concluded that the total teperature should remain constant, too. However, for other refrigerant, this assumption might not work. But still the total enthalpy should be constant in the diffuser, as I know.

[maebr]

Hi again,

I went through the run processing to find if sth is wrong there or there is any unreasonable assumption which results in dropping the total temperature in the diffuser and the following duct. I found a notice at the first of running mentioned in the following:

|The Wall Heat Transfer Coefficient written to the results file for |
| any turbulent phase with heat transfer is based on the turbulent |
| wall function coefficient. It is consistent with the Wall Heat Flux|
| the wall temperature, and the Wall Adjacent Temperature |
| (near-wall temperature). If you would like it to be based on a |
| user-specified bulk temperature instead, please set the expert |
| parameter "tbulk for htc = <value>".|

Considering that I set all the walls to be adiabatic, it should result in the constant total temperature in the diffuser and its following duct. Furthermore I checked and I found that after finishing of the run, the amount of normalized imbalance of heat energy in the diffuser part (Including the following duct) is so high (165%) which I think it shows sth is wrong here.

| Normalised Imbalance Summary |
+--------------------------------------------------------------------+
| Equation | Maximum Flow | Imbalance (%) |
+--------------------------------------------------------------------+
| U-Mom-Default Domain | 7.9432E+01 | 1.5749 |
| V-Mom-Default Domain | 7.9432E+01 | -0.0251 |
| W-Mom-Default Domain | 7.9432E+01 | -0.5992 |
| P-Mass-Default Domai | 8.3670E-03 | -31.0411 |
| U-Mom-Default Domain | 7.9432E+01 | -0.0788 |
| V-Mom-Default Domain | 7.9432E+01 | 0.0336 |
| W-Mom-Default Domain | 7.9432E+01 | -0.0500 |
| P-Mass-Default Domai | 8.3670E-03 | -8.7198 |
+----------------------+-----------------------+---------------------+
| H-Energy-Default Dom | 1.6453E+02 | 165.0522 | !!!!!!!!
| H-Energy-Default Dom | 1.6453E+02 | 9.9387 |
+----------------------+-----------------------+---------------------+

Please let me know if you have any idea about the possible reasons.

Regards,
Mahsa

[Opaque]

Something is completely off, if your P-Mass imbalance is not 0 or small.

The energy imbalance is the least of your problems if the mass is not conserved.


Would you mind plotting the boundary flow for P-Mass at the inlet, and at the outlet? as you the solution is converging you should see two lines in the plot that become flat at the opposite side of the y-axis since one is negative, and the other is positive.

Is there a pitch change across the interface? If there is, the mass flow across the inlet and outlet are different in magnitude by the pitch ratio at the interface.

Check if the computed pitch ratio is consistent with your model.

[maebr]

Originally Posted by Opaque

Something is completely off, if your P-Mass imbalance is not 0 or small.

The energy imbalance is the least of your problems if the mass is not conserved.


Would you mind plotting the boundary flow for P-Mass at the inlet, and at the outlet? as you the solution is converging you should see two lines in the plot that become flat at the opposite side of the y-axis since one is negative, and the other is positive.

Is there a pitch change across the interface? If there is, the mass flow across the inlet and outlet are different in magnitude by the pitch ratio at the interface.

Check if the computed pitch ratio is consistent with your model.

Yes, we checked it and I create a new monitor to show the difference of the inlet and outlet mass flow (considering the direction) that should reach to 0 when it is converged, completely. The amount of this parameter did not reach to 0 but it reached to -0.003 and stayed almost constant. However, considering the low flow rate of 6.7 gr/s, per passage, it might be still high and results in high amount of P-Mass imbalance.
About the pitch change, no there is not any pitch change across the interface.
Please let me know your opinion about it.

[ghorrocks]

You don't need to create a variable for the imbalance, they are already there. Make a new solver monitor and create a new chart - you will see you have the imbalances in there by default.

The question is whether you should add them as a convergence criteria - based on this thread the answer is probably yes.

[TurBoris]

Hi,

Its been almost 1.5 year after the latest thread, but I'd like to add my comment. Have you checked the mass flow averaged absolute total temperature values at inlet and outlet surface of the diffuser. Since diffuser takes the flow from the impeller outlet where the flow field is really complex. The pressure, velocity and temperature quantities are highly non-uniform at diffuser inlet. Probably your absolute total temperature value changes at diffuser inlet from hub to tip or in tangential direction. This holds for the other vectoral or scalar quantities. There will be a flow migration inside the diffuser and if you check your solution data for example at midspan you may observe that absolute total temperature is not constant and varies. Please check the mass flow averaged values of absolute total temperature at diffuser inlet and outlet. There may be a difference around 1 or 2 Kelvin because of the numerics. However, the quantities should be close since no work or heat is added or removed in diffuser row.

[ghorrocks]

Won't that be covered in the imbalances? Any deviation like you describe will go to the boundary, and then the imbalances will pick it up.

[TurBoris]

Now I have checked former posts upon Glenn's question, and I think there are some convergence issues in the diffuser domain. Continuity and energy imbalances are high. This may stem from the separations in diffuser domain. Since the flow is forced towards an adverse pressure gradient, and varying incidence angles from hub to tip give rise to flow separations in diffusers. Probably he sees a backflow, if he set diffuser exit as static pressure outlet and very close to the diffuser blade. The flow domain and BC's are should be properly set in order to converge these types o flows. Go for a transient analysis if convergence problems still exist even if the setup is OK.
Regarding to my last post, I'd like to share my experience. This was related how I post process a diffuser flow field. If one check the absolute total temperature through a turbosurface, lets say, at 0.5 span. Then it can be seen that the absolute total temperature varies. One can say that it is not physical. However, it is meaningful because, the flow migration (from hub to tip or vice versa) takes place across this plane, and fluid with high energy and low energy mixes in such a complex flow. Thus, mass flow averaged absolute total temperatures at inlet and outlet of diffuser should be checked and they must be equal to each other, since the flow is adiabatic and no work is done through the diffuser.