Multidisciplinary Design Optimisation for the Aerospace Industry. Harvey Thompson, School of Mechanical Engineering, University of Leeds

Transcription

1 Multidisciplinary Design Optimisation for the Aerospace Industry Harvey Thompson, School of Mechanical Engineering, University of Leeds

2 Outline Importance of European Aerospace Industry What is Design Optimisation (DO)? What is Multi-disciplinary Design Optimisation (MDO)? AMEDEO project: MCSA ITNs and Doctoral Training in Europe AMEDEO research: Scientific Work Packages and Example Results Summary IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

3 Importance of European Aerospace Industry Growing demand for air travel Need to reduce impact on environment Aviation Annual 220 Billion Euros for EU economy Provides 4.5 million jobs Innovation as key to preserving competitivenes [The European Commission s Flightpath 2050] Key technology: Multidisciplinary Design Opt. (MDO) IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

4 Importance of European Aerospace Industry European Commission has recognised the strategic importance of the EU s aerospace industry in a number of key strategy documents: EU s R&D Programme ( ) Horizon 2020: a key enabler to stimulating the European economy Europe 2020/Flightpath 2050: a catalyst for growth and skilled employment to meet targets for high-added value growth Advisory Council for Aviation Research and Innovation in Europe (ACARE): advanced manufacturing will contribute to increased resource efficiency... through lighter structures UK has world s second largest aerospace industry IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

5 What is Design Optimisation (DO)? 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 5 University of Leeds: Multidisciplinary Design Optimisation for the aerospace industry. 20 th January 2016

6 What is DO? Design Optimisation (DO) Engineering design in aerospace industry typically requires the solution of optimisation problems with: A number of disciplines (e.g. aerodynamics, heat transfer, stress, acoustics, economics, ) A large number of design variables (things the designer can control) typically 100s or even 1000s causes increase in complexity: the curse of dimensionality Many competing objectives: e.g. lift, weight, safety, lifecycle costs, These results in very large, complex computational problems which require specialised Design Optimisation methods. IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

7 What is DO? Design Optimisation (DO) In principle, can identify the optimal ( best ) designs (e.g. shapes of compressor blades) BUT also need to account for UNCERTAINTY due to: Supplier tolerances Manufacturing errors Variable operating conditions Hence need to account for randomness to achieve robust and reliable design stochastic optimisation methods are very popular IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

8 Role of Approximation in Design Optimisation DO problems in aerospace are very large computational problems (3D flow, heat transfer, etc, ) How do we manage resources efficiently? We cannot possibly do extensive parameter studies with large numbers of design variables Use approximations to understand how system responses (lift, drag, surface temperatures, ) vary throughout the design space Surrogate modelling (also meta- and response surface modelling) approximates system responses using values at certain points in design space Simulation models run at parameters defined by Design of Experiments (DoE) in the parameter space IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

9 Role of Approximations in Design Optimisation Design of Experiments (DoE) Methods Minimise the number of computational simulations needed to extract the maximum possible information from them Is a large field of study in its own right (e.g. full factorial sampling, Box Behnken, space filling DoEs, ) Space-filling DoEs For a given number, N, of simulations space-filling techniques prevent empty areas in design space, leading to very inaccurate approximations. This problem can be addressed using Latin Hypercube techniques IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

10 Role of Approximations in Design Optimisation Latin Hypercube DoEs - design space is subdivided into an orthogonal grid with N elements of the same length per parameter (Cavazutti, 2013). Each element is populated by one and only one DoE point, chosen randomly within each element IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

11 Role of Approximations in Design Optimisation Sub-volumes of design space chosen to avoid spurious correlations or to spread DoE sampling points evenly over the entire design space: E.g. N=10 DoE points and 2 design variables A number of techniques for achieving good spread of DoE points throughout design space Optimal Latin Hypercube techniques IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

12 Role of Approximations in Design Optimisation Use stochastic optimisation method (e.g. Genetic Algorithm to generate DoE points) IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

13 Role of Approximations in Design Optimisation Minimum distance plots, N=120, 2 design variables IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

14 Role of Approximations in Design Optimisation Surrogate Models Use statistical methods to approximate system responses (e.g. lift, drag, surface temperatures,,) using data calculated at the DoE points Big research area in statistics nonlinear regression, radial basis functions, kriging, moving least squares, Kriging, N=10 Actual IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

15 Role of Approximations in Design Optimisation Surrogate Models Kriging, N=20 Kriging, N=50 Actual Once we have an accurate model feed into optimisation algorithm to find optimum (e.g minimum drag, maximum lift, lowest surface temperatures, ) Minimum value IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

16 Balancing Multiple Objectives Multi-Objective Optimisation Aerospace optimisation problems always have competing objectives: lift, strength, weight, cost etc with multiple design variables e.g. mechanical power consumed while cooling metal surface: want to minimise both power consumption and temperature Generate Pareto curve by finding optimal cooling power for specified surface temperature Enables designers to strike the best balance IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

17 What is MDO? What is Multidisciplinary Design Optimisation (MDO)? IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

18 Why do we need MDO? The goal of Multidisciplinary Design Optimisation is to optimize the system as a whole Optimal aircraft? taking into account the interactions between disciplines as well as competing objectives. IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

19 Why do we need MDO? MDO avoids generation of sub-optimal design solutions Example shape optimization of flexible wings w.r.t. drag designing without accounting for wing deflection clearly pointless! Ultimate load (7.6m) 1g flight (3.7m) Jig shape (0m) [Boeing.com] Boeing 787 [airliners.net] Traditional approach: Sequential shape optimization IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

20 Why do we need MDO? Shape optimisation of compressor rotor blade (Polykin, Toropov, Shahpar, 2010) Maximise efficiency of rotor = (Poutlet P inlet ) (γ 1)/γ 1 T outlet T inlet 1 leads to large sweep deformation in blade causing large centrifugal stresses at root of blade blade would fail under operating conditions Single discipline (aerodynamic) optimisation leads to structural failure Solution: include a stress constraint in the problem formulation BUT yields a significantly lower efficiency than aerodynamic analysis IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

21 What is MDO? MDO uses optimisation methods to solve design problems incorporating a number of disciplines All relevant disciplines (aerodynamics, stress, heat transfer, cost, life,,) should be accommodated simultaneously MDO seeks to account for interactions between disciplines and competing objectives MDO became more widely used in the aerospace industry due to greater concerns in US military about lifecycle costs rather than pure performance greater emphasis on manufacturability, reliability, maintainability. The MDO optimum is superior to that from sequential optimisations since it exploits interactions between the disciplines BUT at the price of greater complexity IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

22 What is MDO? MDO will be of increasing importance for aerospace design ACARE2020 Vision: future development of environment-friendly aircraft will be based on a systematic, model-based process where MDO is a key enabling technology Flightpath2050: Europe s Vision for Aviation: by 2050 multidisciplinary design and development tools are used routinely and co-operatively to support a high level of integrated system design There are numerous strategies or architectures for organising the various discipline models within the problem formulation IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

23 What is MDO? MDO architectures Optimisation problems are represented in terms of local design variables (associated only with a specific discipline e.g. aerodynamics) and shared design variables which couple the various disciplines together Need to capture effect of shared variables on the discipline optimisation problems e.g. aerodynamic loads at specific control points for the stress optimisation This is often achieved using surrogate models of the effect of shared variables on discipline optima. e.g. BLISS-2000 Excellent recent review of MDO architectures by Martins & Lambie (AIAAJ, 51(9), September 2013) IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

24 What is AMEDEO? The AMEDEO project? Amedeo Clemente Modigliani (Italian painter ) IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

25 AMEDEO AMEDEO (Aerospace Multidisciplinarity-Enabling DEsign Optimisation) is an Initial Training Network (ITN) Founded and funded by the European Commission Consortium of various industrial and academic partners Provides ESRs a great network in academia and industry Offers early-stage researchers (ESRs) excellent training & mentoring Encourages disciplinary, institutional & cultural exchange The scientific mission Bringing together academia and industry to develop software design tools needed to design next generation of environment-friendly aircraft. IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

26 AMEDEO Initial Training Networks are the EU s gold-standard for Doctoral Training Doctoral students receive excellent scientific and transferrable skills training with a focus on: Career development (CVs, job applications, grant proposals, ) Mobility throughout Europe (working in different countries, secondments, ) Input from non-academic sector (entrepreneurship, software tools, ) Cohort-wide training and support (training at 6 monthly network meetings, ) Gender issues, gender balance Improving attractiveness of research careers excellent working conditions and compensation (generous salaries, training budgets, living allowances, ) IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

27 AMEDEO In UK, EPSRC Centres for Doctoral Training (CDTs) are evolving towards the cohort-based ITN model.e have up to 40 PhD studentships available to UK/EU Within Faculty of Engineering at University of Leeds we have 40 fully funded PhD studentships in our CDTs in: Bioenergy Complex Particulate Products and Processes Fluid Dynamics Integrated Tribology Nuclear Fission Tissue Engineering and Regenerative Medicine IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

28 AMEDEO Bringing together academia and industry to develop software design tools needed to design next generation of environment-friendly aircraft. 13 ESRs AMEDEO Associate Partners Airbus Operations France, SFE GmbH Germany Consortium University of Leeds, Rolls Royce, Altair Engineering, Queen Mary London, Von Karman Institute Brussels, Delft University of Technology, Advanced Lightweight Engineering, Technische Universität München, ONERA Paris, Koc University Istanbul IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

29 AMEDEO Research SWP 3: Application of advanced MDO methods to aircraft engine design SWP 1: New computational and parametrisation methods for largescale MDO problems AMEDEO Research [SWP: Scientific work package] SWP 4: Novel applications of MDO to the design of composite aeronautical structures SWP 2: Efficient metamodel-based robust MDO frameworks with multiobjective and multi-fidelity capabilities IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

30 AMEDEO Research IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

31 Graphical Processing Units can offer better performance than CPUs Efficient High Performance Computing Techniques for Multi-Disciplinary Optimization Mohamed Hassanine Aissa 1 Dr. Tom Verstraete 1 Prof. C. Vuik 2 GPUs take instructions from CPUs to bomachinery render images Department onto the screen, Von at Karman a Institute, Belgium stitute very high of Applied bandwidth. Mathematics, This TU Delft, the Netherlands computational power can be beneficial for other uses. 36 th Month Network Meeting, ONERA, 5/10/2015

32 GPU Comutational Power Comparison between CPUs and GPUs Huge performance gap between GPUs and CPUs so use GPUs to do optimisation runs BUT requires specialist programming methods Xeon Phi NOTE: GPUs now being used extensively at Leeds for real-time simulation of electronics cooling (Summers/Delbosc) and infection control in hospitals (Noakes/Khan) 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 32

33 Application: Compressor Cascade Optimisation Nasa.gov Mohamed optimised shape of cascade 23 x faster with GPU than with CPU (5 days compared to 4 months!) 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 33

34 Node-based shape optimization and its application to problems from aerospace industry. Daniel Baumgärtner, Reza Najian Asl, Kai-Uwe Bletzinger Chair of Structural Analysis Technical University Munich

35 Node-based optimisation Current representations in terms of design variables can lead to conservatism restricting innovation in early (preliminary design stages) Goal of node-based optimisation is to enable positions of each node in the computational solution to be a degree of freedom much more scope for design innovation Preliminary aircraft nodebased parametrisation [Martins, 2013] Example: Parameterized aerostructural shape optimization 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 35

36 Node-based optimisation Aero-structural parametrisation of airfoil node-based parametrisation can handle much larger deformations than conventional parametrisations. Pseudo-elasticity inlcuding shear effects (FEM, structured & unstructured meshes) 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 36

37 Node-based optimisation Aero-structural parametrisation of airfoil node-based parametrisation can handle much larger deformations than conventional parametrisations. Pseudo-elasticity inlcuding shear effects (FEM, structured & unstructured meshes) 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 37

38 Node-based optimisation Regularity of computational mesh very important for solution accuracy Efficient mesh regularization to preserve mesh quality during displacements is crucial Wing example: regularization of 150,000 mesh elements in less than 1 minute 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 38

39 Node-based optimisation Regularity of computational mesh very important for solution accuracy Efficient mesh regularization to preserve mesh quality during displacements is crucial Wing example: regularization of 150,000 mesh elements in less than 1 minute 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 39

40 Aircraft Engine Design Fan: draws large volume of air into engine and accelerates it. Typically 90% of air passed directly into jet nozzle to create thrust and 10% to core engine. Compressor: core engine air compressed by rapidly rotating disks ~ 1/50 th of normal volume Combustion: compressed air mixed with fuel which expand to create high speed, temperature (~2300oC) gas Turbine: extract energy from hot gas to drive fan and compressor. Also creates additional thrust Projects with Rolls Royce increasing the efficiency of the turbine design through improved cooling Copyright 2016 Rolls-Royce plc 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 40

41 TURBINE STATOR WELL OPTIMISATION Julien Pohl Gary A. Clayton Harvey Thompson Copyright 2016 Rolls-Royce plc

42 Stator well cooling optimisation Hot gas ingestion Cooling air Cooling air Gas Turbine Engine Typical turbine stator well Copyright 2016 Rolls-Royce plc 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 42

43 Stator well cooling optimisation Cooling problems in gas turbine engine cavities adjacent to the main gas path Thermodynamical Issues Cooling flow bled from compressor (up to 20% overall) Reduction of engine/stage efficiency Direct impact on thermodynamic cycle performance Thermal Issues Hot gas ingestion in cavities Excessive metal temperatures Stresses which reduce component life Cooling air Hot gas ingestion Cooling air Copyright 2016 Rolls-Royce plc 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 43

44 Stator well cooling optimisation Stator Well Flow Structures Baseline Geometry Use deflector plate to guide cooling air towards critical disc surface Deflector Plate Geometry Copyright 2016 Rolls-Royce plc 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 44

45 Stator well cooling optimisation Optimisation results Response Surfaces Results for minimal coolant: Reduction by nearly 70% Most of design points at border of design space Deflector tip now as close as possible to the rotor disc Deflector moved close to stator foot Length of middle curve nearly no influence on performance Copyright 2016 Rolls-Royce plc 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 45

46 Stator well cooling optimisation Optimised Design Baseline Design Optimised deflector plate reduces cooling mass flow rate by 70% whilst meeting cooling requirements (even though hot gas injections increases slightly) Copyright 2016 Rolls-Royce plc 3rd ECCOMAS Young Investigators Conference Aachen 2015 AMEDEO Project 46

47 AMEDEO Research Other research projects are developing MDO methods for: Composite wings, fan blades, fuselages Acoustic vibration in composite fuselages Crashworthiness assessments Surrogate modelling for large optimisation problems IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

48 Summary The aerospace industry is vital to the EU economy and manufacturing sectors Multi-disciplinary Design Optimisation will be a key technology in enabling it to meet its innovation challenges to produce more energy-efficient and sustainable aircraft MDO now routinely used in aerospace design but there are key skills shortages in Europe Key bottlenecks in improving the effectiveness of MDO design processes include Long computer simulation times Need for greater design freedom (large number of design variables) Better exploitation of new materials, e.g. composites AMEDEO is producing cohort of young researchers to drive uptake of MDO IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January

49 Questions? amedeo-itn.eu twitter.com/amedeo_itn facebook.com/amedeoitn IMechE 3rd University Young of Investigators Leeds: Multidisciplinary Conference Aachen Design 2015 Optimisation AMEDEO for Project the aerospace industry. 20 th January