Re effect on domain scaling, mass and area averaging - quasi-3D and 3D domains

Sasquatch · October 26, 2016, 06:28

Hi all,

I didn't find a thread which treats such issue.

I am simulating a flow over a 10 cm long, 3 stage, ORC radial outflow turbine.

I would like to estimate the profile and endwall losses - hence I computed entropy loss coeff for quasi-3D and for 3D.

Quasi-3D domains are 2 cells span-wise and 3D blades do not involve tip leakage. Quasi-3D domain has a free-slip condition on the hub and shroud, 3D has of course no-slip.

Surprisingly, entropy loss coeff. is larger for quasi-3D (without endwall loss).
The shock is also slightly stronger on quasi-3D.

For the preliminary estimation I define my loss coeff. (Denton, 1992 or Coull 2016) as eta = (s_2 - s_1)*T_2/(h_02 - h_2) at inlet/outlet of the domain.

Can my discrepancy originate from the Re scaling effect? Blade dimensions are not scaled.

I am using mass flow averaging. However, using mass flow averaging for quasi-3D and area averaging for 3D cases gives larger loss for 3D due to secondary effect but I think I should use the same averaging to compare them.

For now it looks like my endwall losses are negative. The paper mentions that for low aspect ratios, endwall losses can be negative from such estimation and it is a complex problem but how to tackle that?

Sasquatch · October 27, 2016, 15:42

Does anyone have any hint?

Sasquatch · October 30, 2016, 17:38

I think I have an answer to my question and it is connected to the flaring angle of the domain. Quasi-3D case here is not a good idea to investigate profile losses due to neglecting a certain velocity component which in acts in 3D domain. Hence 3D case without tip clearance and free-slip at the hub and shroud represents the geometry for profile loss investigation. Entry entropy loss coeff. estimation showed some coherence.

October 26, 2016, 06:28	Re effect on domain scaling, mass and area averaging - quasi-3D and 3D domains	#1
Sasquatch Member Join Date: Jul 2016 Posts: 33 Rep Power: 10	Hi all, I didn't find a thread which treats such issue. I am simulating a flow over a 10 cm long, 3 stage, ORC radial outflow turbine. I would like to estimate the profile and endwall losses - hence I computed entropy loss coeff for quasi-3D and for 3D. Quasi-3D domains are 2 cells span-wise and 3D blades do not involve tip leakage. Quasi-3D domain has a free-slip condition on the hub and shroud, 3D has of course no-slip. Surprisingly, entropy loss coeff. is larger for quasi-3D (without endwall loss). The shock is also slightly stronger on quasi-3D. For the preliminary estimation I define my loss coeff. (Denton, 1992 or Coull 2016) as eta = (s_2 - s_1)*T_2/(h_02 - h_2) at inlet/outlet of the domain. Can my discrepancy originate from the Re scaling effect? Blade dimensions are not scaled. I am using mass flow averaging. However, using mass flow averaging for quasi-3D and area averaging for 3D cases gives larger loss for 3D due to secondary effect but I think I should use the same averaging to compare them. For now it looks like my endwall losses are negative. The paper mentions that for low aspect ratios, endwall losses can be negative from such estimation and it is a complex problem but how to tackle that?

October 27, 2016, 15:42		#2
Sasquatch Member Join Date: Jul 2016 Posts: 33 Rep Power: 10	Does anyone have any hint?

October 30, 2016, 17:38		#3
Sasquatch Member Join Date: Jul 2016 Posts: 33 Rep Power: 10	I think I have an answer to my question and it is connected to the flaring angle of the domain. Quasi-3D case here is not a good idea to investigate profile losses due to neglecting a certain velocity component which in acts in 3D domain. Hence 3D case without tip clearance and free-slip at the hub and shroud represents the geometry for profile loss investigation. Entry entropy loss coeff. estimation showed some coherence.