Non Axisymmetric Hub Design Optimization for a High Pressure Compressor Rotor Blade Vicky Iliopoulou, Ingrid Lepot, Ash Mahajan, Cenaero
Framework Target: Reduction of pollution and noise Means: 1. Advanced combustion 2. Heat management 3. Active (injection/aspiration) and Passive control ( SP 5.2 ) Non axisymmetric endwalls efficiency surge margin Loss mechanisms in turbomachines Losses reduction 1. Tip clearance flow 2. Profile losses Blade shape design 3. Secondary flows Circumferential grooves Squealer tip 3D Non axisymmetric endwalls design 3D CFD based optimization
Non axisymmetric endwall profiling Potential impact on the flow field: Local effect close to the endwall: Introduction of some additional curvature on the endwall Influence on the pressure field (e.g. reduction of the crossflow). More «global» effects Change of cross sections Influence on the whole flow field over the span. Modification of the blockage, in particular when steep bumps are applied in the leading edge region.
Layout of the presentation Endwall profiling Parameterization CAD integration Optimization specification Optimization methodology Optimization results Assessments on the fillet impact Conclusions ASME Turbo Expo 2008, GT2008-50293 12AIAA/ISSMO 2008, AIAA-2008-5881 Joint Snecma/Cenaero Patent
Hub parameterization CATIA v5 R17 16 parameters Series of B-spline curves Design between LE and TE 6 main control points in the blade channel that can move radially, axially and/or circumferentially 3D surfaces that follow the blade curvature
Direct CAD access Master script (Python or C++, called by the optimizer) CAD Linux Windows Ref Mesh CAD Mod CAD Mod Ref Mesh Mesh middleware client Unstructured mesh generation TCP / IP Structured mesh generation middleware server CATIA V5 SolidWorks UG NX - Pro/Engineer Open CASCADE -
Optimization specifications Geometry (Fixed) single mobile row CATIA v5 parameterized hub endwall A posteriori profiling Per individual 2 operating points computed: 1 close to peak efficiency and 1 close to the stability limit ( 2.2 M. grid points / tip clearance modeling / RANS k-l Smith turbulence model) Specification 1st Mono-point optimization to freely search the design space Maximize isentropic efficiency (free of constraint) Two-point optimization Maximize isentropic efficiency at design point Constraint on Total-to-Total pressure ratio at close to stall point Manufacturing constraints - Mass flow/outlet angles
Optimization chain set up Maximize efficiency @ design Preserve pressure ratio @ stall Manufacturing constraints Design of Experiments Approximate Models Auto-adaptive RBF networks Evolutionary MO Optimization DATABASE Minamo elsa RANS Post treatment utilities END Performance check Success switch ONLINE modeling Infill criteria
Optimization convergence history Step 2 - Optimization Full stabilization after 40 design iterations Step 1 - DoE 67 converged cases out of 100 individuals/samples
Total pressure ratio constraint at stall Leave-One-Out Cross-correlation coefficient history Key issue of infill Exploitation/Exploration
Overall performance results First mono-point optimization highlighted a marked total pressure drop close to stall Need for robust multi-point design Two-point design: Performance gain at the design point 2 distinct families identified Efficiency increase by 0.4 % Mass flow increase only by 0.4% (DoE scatter > 1%) Total-to-total pressure ratio preserved close to stall Very moderate outlet flow angle alteration Gain should be preserved in a stage environment Checked and confirmed (3D RANS simulations)
Total relative pressure downstream at design point Optimized design Marked losses decrease almost until 50% Axisymmetric reference Local (low mass flow BL zone) losses increase
No reduction of the secondary flows Axisymmetric reference Impingement of the secondary flow at the blade SS No significant change of the structure of the secondary flows Optimized design Additional recirculation zones at wall with non axi surface Local increase of losses
Relative Mach number at the B2B plane at 23.6 % span at design point Optimized design Axisymmetric reference Visible reduction of the wake Marked decrease of the relative Mach number downstream the shock, in the region of flow acceleration
3D contouring impact on the static pressure distribution around the blade 3.5 % span at design point Mono - point Multi - point Local increase of losses close to the wall due to the shock strengthening
3D contouring impact on the static pressure distribution around the blade 23.6 % span at design point Mono - point Multi - point Efficiency improvement due to the reduction of the acceleration downstream the shock
Truncated fillet construction A robust truncated fillet construction was developed under CATIA v5 allowing to maintain O4H default mesh topology in AutoGrid5, essentially relaxing fillet/blade tangency. This truncated fillet construction was constructed in order to ensure robustness w.r.t. large parametric variations of the non axisymmetric platforms. Aerodynamic impact of this truncated fillet was checked against full fillet computations, accounting for bi-tangency through adequate grid topology. Thanks to its computational cost, similar to classical simulation without fillet, and robustness, this truncated fillet construction is exploitable within the design loop so as to directly integrate the fillet impact into the design intent. In the present framework, performances gains obtained through 3D profiling appeared a posteriori preserved with fillet (shock structure alteration).
Integration of real geometry effects Key issue of discretization and model level required
Gain preserved when adding the fillet Relative Mach number at the B2B plane at 45% span AXI No fillet Non AXI No fillet AXI Fillet Non AXI Fillet
Conclusions The 3D endwall hub contouring of the HPC improved the isentropic efficiency by 0.4% at design point, while preserving the total pressure ratio at stall. Local modification but 3D impact. Main loss mechanism (shock-acceleration formation) tackled. The non axisymmetric hub decreased the acceleration downstream the shock and played no evident role on the secondary flows. Two-point 3D hub endwall optimization tackled the main loss mechanism in a similar way while preserving total pressure ratio near stall and offering 2 different families of designs yielding a similar performance gain. Real geometry effects integration (fillet handling): Axi/non-axi gain preserved when fillet modelled.