That's the whole point. There are two approaches:
- The parameter option (your current choice) which needs a load of models to then give a response surface. You then find the "best" solution and refine the surface to work out the parameters. Advantage is you see all of the design space, disadvantage is it's lots of runs.
- Adjoint and morphing move the Fluent mesh to get the "best" result. You don't generally see the rejected designs. Advantage, it's computationally very efficient; disadvantage you don't see the response surface and poor choice of guidelines can result in some interesting results.