In this chapter, the setup and use of the adjoint solver and the mesh morpher/optimizer in Ansys Fluent are described. Both tools offer shape optimization capabilities, but the scope is complementary. The fundamental difference in the technology is that the adjoint solver uses gradient-based optimization by calculating derivatives, whereas the mesh morpher/optimizer uses a strategy that is not based on gradients to decide design updates.
The tools have a common approach to change a design by morphing the computational mesh. This has the benefit that you need not return to the original geometry in order to update the design for the purposes of the optimization.
The two key considerations when deciding whether to use the adjoint solver or the mesh morpher/optimizer are as follows:
Is the flow physics / observable of interest supported by the adjoint solver? (For details about adjoint capabilities, see the following sections: Basic Assumptions and Consistency Checks and General Observables) If not, then the mesh morpher/optimizer is the better option, since there are very few limitations in the supported physics and choice of quantity of interest with this approach. It is important to verify that the quantity of interest can be specified as an adjoint solver observable.
Is the dimension of the design space high? Detailed shape optimization of 3D geometries can involve design spaces with large dimensions (thousands of degrees of freedom or more). The adjoint solver is very effective in this scenario, whereas the use of the mesh morpher/optimizer becomes computationally prohibitive when the number of degrees of freedom exceeds more than, say, 10–20.