This presentation was held at the 2020 NAFEMS UK Conference "Inspiring Innovation through Engineering Simulation". The conference covered topics ranging from traditional FEA and CFD, to new and emerging areas including artificial intelligence, machine learning and EDA.

Resource Abstract

Computational Aero-Acoustics (CAA) is increasingly important across a range of engineering sectors including aerospace and automotive to reduce and manage environmental noise pollution and to enhance the customer experience of new vehicles – particularly electric cars.  

Within aerospace, acoustic engineering is focused on parts of the airframe known to generate noise – including the landing gear.   Current methods for predicting noise sources and propagating the sound make use of either semi-empirical methods or fully scale-resolving simulations.  Semi-empirical methods provide rapid noise predictions however are of limited scope.  Full scale resolving simulations can often require several days or weeks of run-time on large scale supercomputing facilities.  The timescales for simulation turnaround are prohibitive for inclusion in the early design stages.

Zenotech has recently developed a Fast Randomised Particle Mesh method (FRPM) - a CFD-based stochastic broadband acoustics technique that has been shown to be significantly faster than full scale-resolving methods.  

We present two open test cases from the international BANC workshop on airframe noise computations, which are excellent benchmarks.  The first test case “LAGOON” (LAnding Gear nOise database and CAA validatiON) is a relatively simple geometry with complex physics.  The second test case is a complex test case with lots of geometric features called the “PDCC NLG” (Partially-Dressed Closed Cavity Nose Landing Gear).  This case is representative of the type of geometry that we would expect from aerospace landing gear design processes.  

The toolchain includes a RANS step that can either use existing RANS CFD data in a variety of common formats, or generate new data.  The stochastic source generation is very rapid, and can be run in a fraction of the time for the RANS solution.  Propagation of the noise sources is via a high order CFD method that can run on modern GPUs.  The results are shown to be as accurate as current scale resolving methods and are 10x - 100x faster to produce.