An Experimental and Numerical Investigation of the Rotordynamic Coefficients of a Gas Labyrinth Seal: Part 2 - Numerical Simulation

Part 1 of this paper presented the experimental investigation of the rotordynamic behavior of a six-tooth gas labyrinth seal. In part 2, the rotordynamic force coefficients and leakage flow of the labyrinth seal were predicted using a bulk-flow model (BFM) and a computational fluid dynamics (CFD) model. The CFD model solved the Reynolds-averaged Navier-Stokes (RANS) equations with a Spalart-Allmaras turbulence model. A grid independence study was conducted in both the meridional plane and circumferential direction using four different node configurations for both studies. A grid that offered a good compromise between computational cost and solution accuracy was determined. The quasi-steady state method was employed using combinations of four different subsynchronous precessional frequency ratios (PFR) to find the rotordynamic coefficients. Additionally, the shaft was statically displaced 5, 10, and 20 percent of the radial clearance and it was found that the solution did not vary with displacement. A 10 percent displacement was chosen to conduct the rest of the numerical simulations. The comparison between the experimental and computational results showed a good agreement for the rotordynamic coefficients that did not vary with frequency. However, when frequency dependence behavior was observed, the direct damping coefficient was underpredicted by both the bulk-flow model and CFD. Furthermore, the leakage was overpredicted by the numerical results. Because the direct damping coefficient was underpredicted and the leakage was overpredicted, the numerical results were considered conservative. It was also found that refining the mesh improved the agreement between experimental and CFD results, underscoring the importance of mesh choice in accurately modeling the seal’s behavior. Finally, the CFD prediction outperformed the bulk-flow model, but the computational cost of the CFD simulations far exceeded that of the bulk-flow model.