Prediction of Friction Factors and Leakage for Hole Pattern Seals Using Computational Fluid Dynamics and Artificial Neural Network

Abstract

Hole pattern seals (HPS) reduce leakage and suppress rotordynamic instability in high-performance compressors. A bulk flow modeling of HPS to obtain stiffness, mass, and damping coefficients for the HPS requires a friction factor model. This is typically obtained experimentally in a flat plate tester with flow between a smooth and a roughened flat plate. This article presents an alternative approach to training an artificial neural network (ANN) in conjunction with computational fluid dynamics (CFD) modeling to predict friction factors and leakage in a round hole pattern seal. CFD is used to predict friction factors for a large number of round hole pattern flat plate tester configurations. The CFD results are validated by comparison with experimental results for gas and liquid flat plate HPS cases. An ANN is trained using this large dataset of CFD friction factor results. The ANN-predicted friction factors are shown to accurately predict the friction factors as compared with CFD models. These friction factor predictions are then utilized to obtain the Hirs and Moody friction factor coefficients, and subsequently the seal dynamic coefficients.