Models for Predicting Transonic Flutter of a Wing-Section with Sloshing in an Embedded Fuel Tank

TL;DR

This paper develops and compares three computational models, including surrogate models, to predict transonic flutter of a wing with internal fuel sloshing, highlighting the efficiency and accuracy of neural network-based surrogates.

Contribution

It introduces a novel RBF-NN surrogate model for efficient and accurate prediction of fuel sloshing effects on wing flutter, outperforming linearized models.

Findings

RBF-NN surrogate closely matches high-fidelity results

Linearized EMS model limited in nonlinear scenarios

RBF-NN offers low computational cost and high accuracy

Abstract

The present study focuses on the development, application, and comparison of three computational frameworks of varying fidelities for assessing the effects of fuel sloshing in internal fuel tanks on the aeroelastic characteristics of a wing section. The first approach uses the coupling of compressible flow solver for external aerodynamics integrated with structural solver and incompressible multiphase flow solver for fuel sloshing in the embedded fuel tank As time-domain flutter solution of these coupled solvers is computationally expensive, two approximate surrogate models to emulate sloshing flows are considered. One surrogate model utilizes a linearised approach for sloshing load computations by creating an Equivalent Mechanical System (EMS) with its parameters derived from potential flow theory. The other surrogate model aims to efficiently describe the dominant dynamic characteristics of the underlying system by employing the Radial Basis Function Neural Networks (RBF-NN) using limited CFD-based data to calibrate this model. The flutter boundaries of a wing section with and without the effects of fuel sloshing are compared. The limitation of the EMS surrogate to represent nonlinearities are reflected in this study. The RBF-NN surrogate shows remarkable agreement with the high-fidelity solution for sloshing with significantly low computational cost, thereby motivating extension to three-dimensional problems.