Axial pump optimization

Mohamed911 · May 15, 2022, 07:28

Hello everyone,

I was working on an axial flow pump for quite some time.

now, I'm trying to optimize its performance (increase the pressure raise, increase efficiency). the problem is that I'm not sure of the method to use to do that since I have 8 variables to optimize and it will take a long time to do the full experimental plan.

any suggestions?

Ps: I'm using Ansys CFX for my CFD simulations

Indra · May 15, 2022, 13:13

Quote:

Originally Posted by Mohamed911

Hello everyone,

I was working on an axial flow pump for quite some time.

now, I'm trying to optimize its performance (increase the pressure raise, increase efficiency). the problem is that I'm not sure of the method to use to do that since I have 8 variables to optimize and it will take a long time to do the full experimental plan.

any suggestions?

Ps: I'm using Ansys CFX for my CFD simulations

Optimisation is usually done by either using the Adjoint Solver or using the DOE in workbench to get response surfaces and other data on a range of design points. You have to decide which way you wanna go.

ghorrocks · May 15, 2022, 19:10

How to optimise a multi-variable design is a complex topic and there are many approaches in the literature. As Indra says, ANSYS has some optimisation tools built into workbench (and some very powerful ones available available at extra cost), but you do not have to use their tools if you don't want to.

So have a look into workbench's parameterisation tools for basic capabilities, or the optimisation module for advanced stuff. But also look at the literature for the techniques other researcher use.

Mohamed911 · May 16, 2022, 03:36

I tried using the parametrization tool. however, BladeGen doesn't have that option. I also tried using Taguchi's DOE to minimize the number of conducted simulations, but the accuracy of the results is low.

ghorrocks · May 16, 2022, 04:01

What do you mean the accuracy of the results is low? Are the simulations inaccurate, or the optimisation method is not finding a good optimum? If you cannot find a good optimum then try a different method, or pick a few points yourself to define a bit more of the response surface before you start the formal optimisation method.

Mohamed911 · May 16, 2022, 05:44

it's not about the simulations. but Taguchi's DOE reduce the number of conducted simulations significantly (from 4374 to 18). thus, when I wanted to find the objective function, I needed to do a regression of the found data, which is inaccurate since the number of experiments is significantly reduced.

as for another method, till now I couldn't find any.

May 15, 2022, 07:28	Axial pump optimization	#1
Mohamed911 Member Join Date: Dec 2019 Posts: 31 Rep Power: 6	Hello everyone, I was working on an axial flow pump for quite some time. now, I'm trying to optimize its performance (increase the pressure raise, increase efficiency). the problem is that I'm not sure of the method to use to do that since I have 8 variables to optimize and it will take a long time to do the full experimental plan. any suggestions? Ps: I'm using Ansys CFX for my CFD simulations

May 15, 2022, 19:10		#3
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,719 Rep Power: 143	How to optimise a multi-variable design is a complex topic and there are many approaches in the literature. As Indra says, ANSYS has some optimisation tools built into workbench (and some very powerful ones available available at extra cost), but you do not have to use their tools if you don't want to. So have a look into workbench's parameterisation tools for basic capabilities, or the optimisation module for advanced stuff. But also look at the literature for the techniques other researcher use. __________________ Note: I do not answer CFD questions by PM. CFD questions should be posted on the forum.

May 16, 2022, 04:01		#5
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,719 Rep Power: 143	What do you mean the accuracy of the results is low? Are the simulations inaccurate, or the optimisation method is not finding a good optimum? If you cannot find a good optimum then try a different method, or pick a few points yourself to define a bit more of the response surface before you start the formal optimisation method. __________________ Note: I do not answer CFD questions by PM. CFD questions should be posted on the forum.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
axial piston pump simulation in Fluent	Silence	FLUENT	4	October 14, 2022 00:02
Pump power for hoses feeding to nozzles	Rob Wilk	Main CFD Forum	0	February 24, 2021 09:37
Flow simulation in Axial Pump Impeller	aungkyaw	Main CFD Forum	1	December 9, 2014 02:02
CFD-Post axial pump torque	m4D	CFX	0	October 15, 2012 07:48
axial flow pump, inflow boundary	tma	CFX	3	September 26, 2002 15:28

May 16, 2022, 03:36		#4
Mohamed911 Member Join Date: Dec 2019 Posts: 31 Rep Power: 6	I tried using the parametrization tool. however, BladeGen doesn't have that option. I also tried using Taguchi's DOE to minimize the number of conducted simulations, but the accuracy of the results is low.

May 16, 2022, 05:44		#6
Mohamed911 Member Join Date: Dec 2019 Posts: 31 Rep Power: 6	it's not about the simulations. but Taguchi's DOE reduce the number of conducted simulations significantly (from 4374 to 18). thus, when I wanted to find the objective function, I needed to do a regression of the found data, which is inaccurate since the number of experiments is significantly reduced. as for another method, till now I couldn't find any.