1. HPC Midlands Launch: Loughborough Rolls-RoyceCollaborationLeigh LapworthGary PageChief Design SystemsAeronautical & AutomotiveArchitect EngineeringRolls-Royce PlcLoughborough University

2. Introduction Background to Loughborough Rolls-RoyceUniversity Technology Centre (GJP) Rolls-Royce System Design and Future Directions(LL) Loughborough CFD Examples (GJP) Future Directions (LL)2 HPC-Midlands Launch Event 20 March 2013 3. University Technology Centres Rolls-Royce policy to focus academic research withselected University partners Established Global network of University TechnologyCentres (UTCs) Long term research relationships coveringfundamental research through to engine projects3HPC-Midlands Launch Event 20 March 2013 4. UTC Global Research NetworkPerformance in a SeawayTrondheim Hydrodynamics Lightweight StructuresHigh-Mach PropulsionChalmers, Gothenberg and Materials Dresden Purdue, USA Heat Transfer and Multidisciplinary Process Turbines Aerodynamics Power Electronics Centre IntegrationCottbusMadrid (ITP) Xian ChinaVirginia Tech Thermal Management Pusan, KoreaFuel Cell SystemsTurbines and Combustion Genoa DarmstadtAerospace Materials NIMS JapanDesign partnershipCambridge, Sheffield, SouthamptonSolid Mechanics OxfordAdvanced TechnologySystems & Software Engineering CentreSingapore YorkGas Turbine TransmissionSystems NottinghamControl & Systems EngineeringSheffieldMaterials PartnershipManufacturing TechnologyCambridge, Birmingham & SwanseaNottingham Materials Damping Technology Vibration SheffieldUniversity Gas Turbine Partnership (UGTImperial College Cambridge Electrical Power SystemsPerformanceStrathclydeAdv. Manufacturing Research CentreResearchCranfield Sheffield AMRCcentres Combustion SystemAdvanced Electrical Machines and Drives Aerothermal Processes SheffieldLoughboroughUTCsHeat Transfer and Aerodynamics Electrical Systems for Extreme EnvironmentsOxford (Osney lab) ManchesterNoiseThermo-Fluid SystemsSouthampton Surrey 5. University Technology Centres Loughborough UTC in Combustion Aerodynamicscreated in 1991 (third to be established) In 1991: 3 academics, 2 technicians, 5 researchers In 2013: 7 academics, 4 technicians, 1 expt officer, 1administrator, 5 senior research staff, 19 researchers Typically around 1.6M per year turnover with 40ongoing projects Name updated to Combustion System Aerothermal Processesto reflect increased interest in heat transfer,acoustics, two-phase flow5 HPC-Midlands Launch Event 20 March 2013 6. Combustion Systems ~75% projects on combustion system ~25% other: compressor, ducts, military nozzle, jet noise,instrumentation Experimental and Computational studies (Computational FluidDynamics)6 HPC-Midlands Launch Event 20 March 2013 7. Computational Fluid Dynamics ResearchAccess to RR corporate CFD codes Hydra and Precise as well as associated grid generation (Padram) and visualisation (ss02) Access to source code where required Develop capability (e.g. Hydra Large Eddy Simulation) that is incorporated back into production software Good network with CFD researchers from other UTCs Royal Society funded secondments of RR staff to Loughborough (Leigh Lapworth and Indi Tristanto) Focus on pushing the boundaries of simulation 7HPC-Midlands Launch Event 20 March 2013 capabilities using HPC 8. 8 Simulation in Design Rolls-Royce uses a wide variety of analysis techniques fordesign verification from whole engine to component, manyrelying on HPCComputational fluid Cost modelling is useddynamics is used to to identify cost drivers Finite element structuralunderstand theand maximise value analysis is used foraerodynamics of enginesvibration, lifing andin order to maximise thermals analysis, bothperformance. linear and non-linear at component and sub- system level.Multi-disciplinary WholeEngine Design Systemspredict the behaviour ofthe integrated product Combined CFD and Structural analysis is used to study forced vibration on turbomachineryMaterials designed forrequired properties 2013 Rolls-Royce plc 9. Impact of HPC Simulating the stability boundary Steady multistage CFD performs well at the design point, but cannotpredict stability boundary. Unsteady multistage CFD with sliding planes performs much betternear the stall boundary. 0.69085 0.5Pressure Rise Coefficient, y80 Efficiency, h 0.475 0.3 EXP UNSTEADY 70 STEADY EXP 4 stage research0.2 UNSTEADY 65 STEADY compressor 0.1600.38 0.43 0.480.53 0.58Vx/Umid 2013 Rolls-Royce plc 10. Our HPC Ambitions10 Component Level Increased component capabilitiesSub-system Level Reduced component unit cost(e.g. Integrated HP Spool) Reduced component life-cycle costMulti-component Multi-physics Multi-fidelity Optimised sub-systemsRoutine todayImproved Product Performance Reduced Environmental ImpactFull Virtual Engine Virtual Product Systems Optimised & Robust Systems Step change in: Some routine today Through life operational & Some less frequent environmental performance Number & range of complexproduct systemsThe future 2013 Rolls-Royce plc 11. Compressor CFD Simulations(Andrew McMullan, RR/TSB)Modern research compressor at Cambridge UniversityUnsteady Large Eddy Simulation to resolve all but the smallest scales of turbulenceRR Hydra CFD codeNeed 30 million grid points per blade giving total of 156 million points,grid file 26 GBrun on 360 cores using 150GB memory for >1 month11 HPC-Midlands Launch Event 20 March 2013 12. Compressor Flow Visualisation12HPC-Midlands Launch Event 20 March 2013 13. Prediction of High Speed Jets(Tim Coates, EPSRC CASE, RR Bristol) Mesh for CFD High speed jet rig in Steady RANS andUTC high pressure Unsteady LESfacility Current LES 150 milliongrid points using 800 cores on Hera13HPC-Midlands Launch Event 20 March 2013 14. Unsteady LES Flow Visualisation14HPC-Midlands Launch Event 20 March 2013 15. Improvement in AccuracyJet Centreline Velocity15 HPC-Midlands Launch Event 20 March 2013 16. Study of Curvature feeding Jet Total pressureMach number 16HPC-Midlands Launch Event 20 March 2013 17. 17CFD Simulation future directions There are 5 strands to the future direction of CFD simulation allrequiring appropriate HPC resources and offering improvedpredictive accuracy. All are required to meet the goal of a truly virtualengine. The strands are:l Increased geometric fidelity in the modell Removal of simplifying (steady state) assumptions and artificial boundaries- Increased use of unsteady CFDl Increased coupling between components and disciplines- Particularly between the CFD and the stress and thermo-mechanicalmodels to get the true running shape of the enginel Software engineering to allow calculations of the order of 1 billion nodes toscale to very large numbers of cores and to be set-up and post processed.l Increased physical fidelity- e.g. transition and turbulence modelling that truly represent reality These need to be supported by geometry models and meshgenerators of commensurate capability and quality. 2013 Rolls-Royce plc