Project Goal: To solve for obstacles that minimize drag using derivative-free optimization methods.
Solving for obstacles which minimize drag is a staple problem in fluid mechanics and engineering. This type of problem can be solved using either analytic or computational methods. This optimization problem can be solved numerically using a variety of methods such as gradient based methods or method of moving asymptotes. Instead of these, I have opted to utilize a mesh adaptive direct search (MADS) algorithm, a method of optimization which avoids calculating derivatives of the objective function. The major downside of this method is the speed at which the solver converges; derivative free methods are quite slow in comparison to their gradient based counterparts. However, using a derivative free method can be quite beneficial in problems where finding the derivative of the objective function comes with a high computational cost. Further, MADS is an extraordinarily flexible algorithm. The solver I have built can be applied to a variety of numerical optimization problems, particularly shape optimization problems.
The initial goal of this project was to determine flow obstacles that generate user specified coherent structures, specifically Lagrangian coherent structures (LCS). Lagrangian coherent structures characterize the local behavior of a fluid parcel, and are commonly thought of as the “skeletal structure” of a fluid flow. They are typically defined as the locally most attracting and repelling regions of a flow. The ability to manipulate the formation and location of these structures could be useful in a variety of flow control applications present in engineering, biology, and oceanography. As a long term goal of this project, I would eventually like to apply my solver to the shape optimization problem of finding the obstacle S in which the flow gives a specified LCS. In particular, for a fixed oncoming flow and desired LCS, what should the shape and location of S be?
Associated Media
Drumm, D. (2024). Using Black-Box Optimization Techniques to Solve Shape Optimization Problems in Fluid Mechanics. [Doctoral dissertation, Worcester Polytechnic Institute]. [Full Text].