Learning Forum

[Jason Sum]

Dear all, 

I am now working on shape optimisation of a single-stage Tesla Valve (as shown below) by using parametric design on Design Exploration. But I found too computationally time-consuming for the numerous combinations, and I am planning to adopt the adjoint method in Fluent. 

As the function of a tesla valve is to minimise flow in one way while maximise flow in another way, thats totally two separate simulation cases required (for foward flow and reverse flow), different B.C. are set up for the two cases. I know that for Parametric Design on Design Exploration (DoE), it allows two cases solved data (foward flow and reverse flow) to be integrated and obtain an overall optimised result as shown below. 

The question is,

(1) how about the Adjoint Method, I saw online that it usually proceed from a single simulation case to the sensitivty analysis and morphing. Does it allow data from two simulation cases (forward and reverse flow), then morph the geometry and optimise to achieve maxmised forward flow and minimised reverse flow? Thanks for the guide.

(2) Any recommendations for my case, which shape optimisation methds are efficient to implement. I found the parametric design method limits the possibility of shape changing as it requires the user to figure out and pre-define the parameters one can think of, which may not be the most optimised geometry. 

[Murari Iyengar]

Hi, you can use the Adjoint Solver if the geometry is the same as Adjoint optimizer supports multiple operating condition optimization. You can refer to the tutorial to get started: Chapter 30: Using the Adjoint Solver – 2D Laminar Flow Past a Cylinder (ansys.com)

[Jason Sum]

Hi Murari, I saw that the tutorial demonstrating the shape optimisation with two output objectives (i.e. Drag and Lift) to be minimised or maximised under the same solving Boundary Conditions.

But what if in my case, the tesla valve geometry same but there are two solving cases (i.e. two ansys solvers required independenly to solve the forward flow case and backward flow case, ), can Adjoint Solver still be used to optimise the geometry shape for maxmising forward flow and minimising backward flow. Thanks and your further help is important to me!

[Jason Sum]

Thanks, as I do not have the Account Number for my student account, could you guide me how could I access the tutorials please!

[Rob]

If you click on Help in Fluent you'll bypass the checks, or try https://ansyshelp.ansys.com/public/account/secured?returnurl=/Views/Secured/main_page.html?hl=f  which is the Public version of the documentation. 

[Jason Sum]

Thanks Rob.

I saw that the tutorial demonstrating the shape optimisation with two output objectives (i.e. Drag and Lift) to be minimised or maximised under the same Boundary Conditions setup on the cylinder.

But what if in my case, the tesla valve geometry remains same but there are two solving cases (i.e. two separate ansys solving processes required to respectively solve the forward flow case and backward flow case), can Adjoint Solver still be used to optimize the geometry shape for maxmising forward flow and minimising backward flow?

I saw that the sensitivity data can be exported and further loaded for morphing. In my case, can the sensitivity data of the forward flow case be exported and loaded into the design tool interface (as shown below) of the backward flow case, so that the both sensitivity data can be integrated together for further shape optimisation (max. forward flow & min. backflow flow). 

 

Thanks and your further help is important to me!

[Rob]

Not sure, I suspect you'll need a more advanced optimisation tool, so may need to the Morpher/Project tools. Have a look at Optislang and Minerva and see what you think: it's a little outside my field of expertise.