A deep artificial neural network assisted genetic algorithm method to optimize a slotted hydrofoil

Abstract

Slotting a hydrofoil is an effective way to passively control the cavitating flow to reduce the cavity pocket size leading to a reduction in vibration, noise, and erosion. However, this comes with losses in the hydrodynamic performance of the hydrofoil including its lift coefficient. To avoid this as much as possible, optimizing the slotted hydrofoil in terms of the location and the angle is of prime importance. The optimization is achieved by designing a deep Artificial Neural Network (ANN) to act as a surrogate model in the process of the genetic algorithm. The training dataset in deep ANN is gathered through simulations that are based on a newly developed Frink finite volume solver for the preconditioned Euler equations. Results obtained indicate that this optimization approach is effective such that the cavity pocket size can be shrieked by about 60 % with a penalty of about 10 % reduction in the hydrofoil’s lift coefficient.