Accelerating Convergence of Fluid Dynamics Simulations with Convolutional Neural Networks

Abstract

A novel technique to accelerate optimization-driven aerodynamic shape design is presented in the paper. The methodology of optimization-driven design is based on the automated evaluation of many similar shapes which are generated according to the output of an optimization algorithm. The vast amount of numerical simulations makes this process slow but the resource-intensive simulation part can be changed to so-called surrogate models or metamodels. However, there isn’t any guarantee of this solution’s accuracy.
The motivation of this work was to develop an acceleration method to speedup optimization sessions without losing accuracy. Accordingly, the numerical simulation is kept in the pipeline but it is initialized with a velocity field that is close to the expected solution. This velocity field is generated by a predictive initializer that is based on a convolutional deep neural network. The network has to be trained before using it for initializing numerical simulations; the data generation, the training and the testing of the network are described in the paper as well.
The performance of predictive initializing is presented through a test problem based on a deformable u-bend geometry. It has been found that the speedup in the convergence speed of the numerical simulation depends on the resolution of the predictive initializer and on the handling of velocities in the near-wall cells. A speedup of 13 % was achieved on a test set of 500 geometries that were not seen by the initializer before.