large scale industrial simulations using hpc · 2020. 1. 16. · modelling, etc. multi-disciplinary...
TRANSCRIPT
Rolls-Royce proprietary information © 2019 Rolls-Royce plc
The information in this document is the
property of Rolls-Royce plc and may not be
copied or communicated to a third party, or
used for any purpose other than that for
which it is supplied without the express
written consent of Rolls-Royce plc.
This information is given in good faith based
upon the latest information available to Rolls-
Royce plc, no warranty or representation is
given concerning such information, which
must not be taken as establishing any
contractual or other commitment binding
upon Rolls-Royce plc or any of its subsidiary
or associated companies.
Large Scale Industrial Simulations Using HPC
Leigh Lapworth Fellow – Computational Sciences Rolls-Royce plc
HPC-AI Advisory Council – 2019 UK Conference Leicester, 16-17 September 2019
WHITE page number and reference – adjust title slide on main master slide No 2
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Rolls-Royce content only Not subject to Export Control
2
Business overview
01
© 2019 Rolls-Royce Rolls-Royce Proprietary Information
3
Rolls-Royce at a glance
Civil Aerospace
£7.4bn
Defence
£3.2bn
Power Systems
£3.5bn
13,000 engines in service around the world
35 types of commercial aircraft powered by Rolls-Royce engines
24,600 employees
16,000 engines in service around the world
Over 150 Customers in over 100 countries
9,800 employees
>20,000 Reciprocating engines sold per year
>1,200 Development, service production, and dealership locations
11,400 employees
£1,378m R&D spend (gross)
Global presence in
50 countries
892 patents approved for filing
© 2019 Rolls-Royce Rolls-Royce Proprietary Information
4
Digitisation
Growing demand for cleaner, safer and more competitive power
Electrification
Our Future Increasing demand for travel, trade and sustainable energy
Fusion of mechanical and electrical technologies
Fusion of physical and digital technologies
As pioneers, we must continuously innovate to provide the best solutions in the markets we serve.
In the coming years, we believe that the three key trends will define the world’s future power needs.
© 2019 Rolls-Royce Rolls-Royce Proprietary Information
5
Digitisation
Cyber Protection
Cyber Protection
Design Manufacturing Services & support
Digital thread
Supply chain
• Virtual Reality, e.g. for design visualization
• Internet of Things, e.g. for parts tracking
• Machine Learning and Artificial Intelligence, e.g. for in-service analytics
© 2019 Rolls-Royce Rolls-Royce Proprietary Information
6
Reinvent with digital
• Refining the Digital Twin
• Improving our productivity
• Generating new service offerings
• Pioneering new possibilities for design
• Broadening markets for Rolls-Royce services
Insert video
WHITE page number and reference – adjust title slide on main master slide No 2
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Rolls-Royce content only Not subject to Export Control
7
Modelling and Simulation
02
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control 8
The challenge of gas turbines
Fuel burns in the Trent engine's combustion chamber at temperatures up to 2,000°C, which is well above the 1,300°C at which some component metals used would start to melt.
BLUE page number and reference – adjust title slide on main master slide No 1
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Rolls-Royce content only Not subject to Export Control
9
Simulation categories
Simulations include computational fluid dynamics, structures and dynamics, impact, combustion, thermo-mechanics, aero-mechanics, aero-acoustics, materials structure, process modelling, etc.
Multi-Disciplinary Design &
Optimization
Search large design space
Produce Best Design
Higher Fidelity Physical
Modelling
Increase accuracy
Improve Understanding
Virtual Product Modelling
Replace product tests
Reduce Cost & Risk
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control 10
The Trent XWB
The most intense and comprehensive development programme ever undertaken by Rolls-Royce.
Six times the computing power applied than the previous generation
https://www.rolls-royce.com/products-and-services/civil-aerospace/airlines/trent-xwb.aspx#section-overview
Compressor: Module weight savings of 15% and
aerodynamic efficiency improvements via the use of compressor blisk technology
Fan: World-beating
levels of performance and
noise with reduced operational cost
Optimised internal air system:
reduces core air demand and reduces fuel consumption
Combustor: proven reliability
that is also cleaner than all current
and future emissions targets.
Turbine: the highest
efficiency turbine system of any Trent engine.
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control 11
The Trent XWB
The world's most efficient large aero-engine
15% Fuel Consumption advantage over the
original Trent engine
$2.9M Savings per year
per aircraft on fuel alone
50,000 Horsepower
generated by 68 high pressure turbine blades
1,600+ Trent XWB
engines on order worldwide
Trent XWB for the Airbus A350 XWB family
https://www.rolls-royce.com/products-and-services/civil-aerospace/airlines/trent-xwb.aspx#section-overview
WHITE page number and reference – adjust title slide on main master slide No 2
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Rolls-Royce content only Not subject to Export Control
12
High Performance Computing
03
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control 13
Computational research and innovation model Work with leading universities and pre-competitive access to national e-Infrastructure to scale codes, develop new physics and design concepts.
Physical Science R&I Validation of codes and
new physical understanding
Break-through test cases
Production systems Proven capabilities
Computational Science R&I Scaling codes to run large
models efficiently
Design Science R&I Use of codes in
optimisation and novel design concept
Future business requirements
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control 14
Developing large scale applications
18-24 months pre-competitive code scaling.
Codes running on new RR production systems from outset.
Trying to do this on a production system is more difficult and capability takes longer to develop with end-users adversely affected.
Archer 2
Archer
Sp
eed
/Cap
ab
ilit
y
Year
Archer pilot +
Leadership Scheme
EPSRC Prosperity
Partnership
In-house
HPC
In-house
HPC
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control 15
Scaling from steady to unsteady CFD
Weak scaling due to many more aerofoils in the simulations.
Strong scaling due to unsteady simulations needing many more time steps than a steady simulations.
Model Type
Mesh Type Run Type Core hours
Steady
Single passage with
O(10) blades
Convergence acceleration
used
O(102)
Unsteady
Full annulus With approx.
O(1000) blades
Tiny time steps for several revolutions
O(107)
>50,000 times increase in computing
times
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control 16
Large scale CFD
Our first ever 2 billion cell simulation.
WHITE page number and reference – adjust title slide on main master slide No 2
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Rolls-Royce content only Not subject to Export Control
17
Future directions
04
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control 18
Our ambition
High-fidelity simulation of a complete gas-turbine engine during operation, simultaneously including thermo-mechanics, electromagnetics, and computational fluid dynamics.
• Challenge: Coupled multi-discipline/scale – not a single code
• Complexity: > 1 trillion cells
• Usage: ~ 1 billion core hours per calculation
• Data files: 10-100 Petabytes per solution
• Energy: 10GWh (using today’s technology)
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control 19
Strategic computational themes
Extreme Computing
• Highly scalable simulations
• Coupled multi-scale and multi-disciplinary codes
• Access to leadership class systems
Trusted Computing
• Verification and validation at extreme scales
• Secure use of 3rd party systems
• Software & data protection
• Software Quality • Skills
development
Lean Computing
• Optimal asset utilisation
• Dashboards • Cost minimisation • Productivity
improvements • Web portals and
apps
Computing Platforms
• Supercomputers • Cloud computing • Emerging
hardware • Visualisation • Augmented and
Virtual Reality • Optimised data
access
Holistic view of factors influencing industrial application of exascale computing.
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control 20
Thermo-Mechanical Grand Challenges
Computational Science
Physical Science
New science and novel designs
Ultra high resolution and extreme scaling
Modelling what cannot be
modelled today
Several orders of magnitude improvement
in computational performance
Trillion cell simulations running on millions of computing cores
Unique new 5-year partnership
• Combining Computational and Physical Science Research.
• Aim is to achieve the world’s first high-fidelity simulation of a complete gas-turbine engine during operation.
• Launched 1st October 2018.
Partners
• Rolls-Royce (lead), CFMS, Zenotech, Universities of Edinburgh (lead), Bristol, Cambridge, Oxford, Warwick.
Associate partners
• Intel, Arm, Microsoft, NCSA.
EPSRC Prosperity Partnership Award
Strategic Partnership in Computational Sciences for Advanced Simulation and Modelling of Virtual Systems (ASiMoV)
Rolls-Royce Proprietary Data | © 2019 Rolls-Royce Not subject to Export Control
©2019 Rolls-Royce