Algorithm Based Generation of Realistic Bristle Distribution: Comparison Between an Unstructured and a Structured Bristle Package for Brush Seal Simulation

Abstract Reducing the leakage in aero engines offers a high potential for efficiency improvements and fuel savings in air transport. Brush seals are able to significantly reduce leakage in comparison to the predominantly used labyrinth seals, but their application in gas turbines is restricted by the limited ability to predict their in-service behavior and leakage rates. This work contributes to enhancing brush seal simulations and improving their predictive accuracy. With the use of image processing algorithms, microscope images from a brush seal test stand are evaluated and individual bristles identified. A statistical analysis on these images determines the distribution of bristles in the bristle package. For pressure differences from 0 to 5 bar across the seal, representative unstructured bristle distributions can then be generated and used as input for CFD simulations. In the simulations, the unstructured bristle package is compared with a structured bristle setup using the same mean inter-bristle distance. The comparison is conducted in 2D as well as 3D. The 3D geometry includes a part of the support ring with a fence height of 0.5 mm and contact between bristles and rotor surface. Simulations with an unstructured bristle package show good agreement with experimental findings. The comparison with a structured bristle package reveals significant differences, when the inter-bristle distance is not varied to fit experimental leakage data. This shows the importance of simulating brush seals using a realistic bristle distribution.