Smart Shape Optimization with ANSYS Adjoint Solver
Hi everyone, here's a challenge that I have for this short video. We're going to see how to best reduce the pressure drop of the exhaust system you're seeing here. The question is, how do we optimize this design to reduce the pressure drop?
Identifying Key Regions
How do we identify the region which will have the most influence on the pressure drop? Would we do physical testing? Well, that sounds pretty expensive and time-consuming. I really want to use simulation, but do I need to define tens of geometrical parameters and perform hundreds of simulations? There must be a smarter and faster solution than that.
Introducing the Adjoint Solver
Of course, there is a smarter and faster solution: the adjoint solver. In only four simulations, we're able to reduce the pressure drop by 12%. Here is a new design.
Key Questions
- How did we get there?
- How do we get visual key information like the area of the geometry that affects performance the most?
- How do we get key information indicating how to modify the geometry for better performance?
- In which direction should we modify the geometry? By which amplitude?
Using ANSYS Fluent 16
Let's see how that's done in ANSYS Fluent 16:
Step-by-Step Process
- Define the Observable: Our goal is to minimize the pressure drop between our four inlets and the outlet.
- Set Up the Adjoint Solver: The adjoint solver is similar to a CFD solver, solving a given set of equations.
- Control Setup: Select auto-adjust controls to ensure the solver converges automatically.
- Initialize and Calculate: Set up residual monitors, initialize the adjoint solution, and calculate it.
- Modify Geometry: Use the results to modify the wall geometry to reduce pressure drop.
- Define Region: Specify which part of the geometry to modify, excluding inlets and outlets.
- Set Region Conditions: Consider symmetry and other boundaries. Input constraints if necessary.
- Numerics Setup: Configure the numerics for the actual change in shape.
- Calculate Design Change: Easily calculate the design change with one click.
Results and Analysis
Now, let's look at the type of result the adjoint solver provides and how it performs a change in shape. The shape sensitivity magnitude with regard to the pressure drop shows that the regions influencing the pressure drop the most are:
- The elbow
- The region where the four inlets join
In this study, we decided to optimize the region where the four inlets join. The adjoint solver has all the information on how the shape of this region impacts the pressure drop, allowing it to make effective changes.
Conclusion
The adjoint solver is a fast and smart shape optimization technique. By following the process four times, we achieved a 12% reduction in pressure drop. From everyone at ANSYS and Ozen Engineering, Inc., thank you very much for your time and attention.
Hi everyone, here's a challenge for this short video. We're going to see how to best reduce the pressure drop of the exhaust system you're seeing here. The question is, how to optimize this design to reduce the pressure drop?
How to identify the region which will have the most influence on the pressure drop? Would we do physical testing? Well, that sounds expensive and time-consuming. I really want to use simulation, but do I need to define tens of geometrical parameters and perform hundreds of simulations?
There must be a smarter and faster solution than that. Of course, there is a smarter and faster solution. It's the adjoint solver. In only four simulations, we're able to reduce the pressure drop by 12%. Here is a new design. How did we get there?
How do we get to visual key information like the area of the geometry that affects performance the most? How do we get key information indicating how to modify the geometry to get better performance out of the design? In which direction to modify the geometry? By which amplitude?
Let's see how that's done in ANSYS Fluent 16. First, we're going to define our observable. Obviously, this is a pressure drop. This is what we want to optimize. We'll define it between our four inlets and the outlet. Then we make sure we're minimizing the pressure drop.
Next, we'll set up the adjoint solver. The adjoint solver is like another solver, for example, like a CFD solver. It will solve a given set of equations. Next, I'm setting up the control of the adjoint solver. I will select auto-adjust controls.
This is the best way to set up the adjoint solver because the solver will automatically adjust to make sure it converges. Next, I set up my residual monitors and initialize the adjoint solution and calculate the adjoint solution. We monitor the simulation and let the adjoint solver converge.
Now, the engineering work starts. We're going to use the results of the adjoint solver and say that we want to modify the geometry of the wall, to reduce the pressure drop. I'm going to apply the target. Then we go into the region and actually define which part of the geometry we want to modify.
Because we don't want to modify the inlet and the outlet, for example. Here is the region we will modify. That's really where we want to get the best design and the best geometry, to make sure we reduce the pressure drop. Then we set up our region condition. Do we have any symmetry?
Do we have any other boundary to take care of? Then we can put some design condition. We're not going to have any here, but design condition is really where you can input a constraint. Constraint on the motion of some of the walls, etc. Then we set up the numerics for the actual change in shape.
And we're all set. We go back to the design change, but now we know what it is to calculate the design change. Which we're doing very easily, just one click of a button.
Now, I showed you how to set up the adjoint solver, how to run it, and how to set up a shape optimization based on the results of the adjoint solver. What I wanted to do now is actually to show you the type of result the adjoint solver is giving us. And how it performs a change in shape.
What you're looking at is the shape sensitivity magnitude with regard to the pressure drop. And what you're seeing is the region that influences the pressure drop the most are A, the elbow, and B, the region where the fore inlet joins.
And what we decided to do in this study is we decided to optimize the region where the fore inlet joins. Here the adjoint solver has the entire information on how the actual shape of this region impacts the pressure drop. Which means it knows how to change it to reduce the pressure drop.
I am sure you now agree that the adjoint solver is very useful. The adjoint solver is a fast and smart shape optimization technique. The process I showed you, do it four times and look at the original geometry and the optimized one with 12% reduction in pressure drop.
From everyone at ANSYS, thank you very much for your time and attention. Thank you.
