Pro Opt Agenda 1 Kort presentation av deltagarna

Pro. Opt Agenda • 1. Kort presentation av deltagarna (Alla) • 2. Presentation av Pro. Opts övergripande ide och mål (Li. U och JTH) • 3. Rekryteringsläget (Li. U och JTH) • 4. Presentationer av doktorandprojektens inriktningar (Li. U och JTH) • 5. Presentation av deltagande företag och deras relation till Optimeringsdriven design (Deltagande företag) • 6. Övriga frågor (Alla)

Pro. Opt Rekryteringsläget - Doktorander • Av de tre ”akademiska” doktoranderna finns ett namn helt klart (Jönköping, Martin Tapankov). Namn på de två Linköpingsdoktoranderna hoppas vi ha klart inom någon veckas tid. • Angående till Pro. Opt relaterade projekt gäller: – NFFP (Saab Aerosystems, mfl): Erik Holmberg – FFI (Saab Automobile, Volvo): Ett namn klart? – FFI (Volvo Truck): Jönköping

Pro. Opt Optimization Driven Design Presentation för Ledningsgruppen 15 Oktober 2009 Linköping GM Powertrain, Kongsberg Automotive, Saab Automobile, Volvo 3 P Altair, Cyb. Aero, ENMESH, ESTECO, Engineering Research Li. U, JTH

Pro. Opt Background • Large scale FEA is an industrial standard today. (even nonlinear) • However, initial designs and cumulative design changes are still very much a matter of experience and intuition only, without computational indications of trends and consequences. Topology opt. CAD FEA Optimization CAD FEA Prototyp Test Respons surface opt. New concepts • Optimization Driven Design (ODD) means that design concepts and re-designs are found as outcomes from precisely formulated optimization problems. → best design within the currently considered requirements.

Pro. Opt in short • Decision support through optimization • We intend to develop both methods and procedures • Pro. Opt leads to better designs faster and cheaper • ”Next generation virtual prototyping” Topology opt. CAD FEA Optimization CAD FEA Respons surface opt. New concept Conceptual and Preliminary Design Detailed Design Prototyp Test

Pro. Opt The Design Process • Conceptual design – basic requirements gives rough choices of material and geometry (pre CAD) • Preliminary design – more precise requirements gives preliminary sizing and material selection in a global picture • Detailed design – final (detailed, local) design based on additional requirements Topology opt. CAD FEA Optimization CAD FEA Prototyp Test Respons surface opt. New concept Conceptual and Preliminary Design Detailed Design A main concern in Pro. Opt is to introduce objectives such as cost of manufacturing and life cycle cost as early as possible in the design process

Pro. Opt Conceptual design • As examples of the conceptual design phase we give some topology optimization solutions

Pro. Opt Topology Optimisation 3 D

Pro. Opt 11/9/2020 A benchmark in 3 D stni@jth. hj. se 9

Pro. Opt Preliminary design • As an example of the preliminary design phase we show an optimization of a product family with respect to size, shape and topology Conceptual design Preliminary design with I-beams

Pro. Optimal Modular Product Families • • • Pre 3 D CAD activity Predefined modules Multiple load cases from prescribed accelerations Different types of design variables Estimate loss of efficiency

Pro. Opt Detailed design • In the detailed design phase metamodels and response surface techniques becomes particularly useful

Pro. Opt Response Surface Methodology Metamodell H or g ROI Direkt metod xi

Pro. Optimization in product realization

Pro. Opt Academic Partners • Professor Anders Klarbring, Linköping University. • Professor Larsgunnar Nilsson, Linköping University • Associate Professor Niclas Strömberg, Jönköping University • Associate Professor Bo Torstenfelt, Linköping University

Pro. Opt Industrial Partners – Major Swedish Industries • • • GM Powertrain, Trollhättan (2580 k. SEK) Kongsberg Automotive, Mullsjö (900 KSEK) Saab Aerosystems, Linköping (1600 k. SEK) Saab Automobile, Trollhättan (3774 k. SEK) Volvo Car Corporation, Göteborg (4716 k. SEK) Volvo Truck Corporation, Göteborg (3000 k. SEK)

Pro. Opt • • Industrial Partners - SMEs Altair Engineering, Lund (900 k. SEK) Cyb. Aero AB, Linköping (900 k. SEK) ENMESH AB, Göteborg (900 k. SEK) ESTECO Nordic AB, Lund (900 k. SEK)

Pro. Opt Industrial involvement • Pro. Opt will start with a kick-of period where a survey of industrial and research needs will be achieved by a workshop as well as by smaller meetings. • This will lead to the formulation of a first generation of demonstrators. • A second generation of demonstrators is to be formulated at a later more mature phase of the Ph. D projects. • We also plan to organize two conferences which will involve industry (also those not directly involved in the project) as well as some invited international researchers.

Pro. Opt 1. 2. 3. Three academic subprojects Multilevel and collaborative optimization Prof. Larsgunnar Nilsson Optimization driven design of machine components Assoc. Prof. Niclas Strömberg Cost and game theory in multiobjective and multiphysics optimization Prof. Anders Klarbring These three projects are strongly connected and the Ph. D students, together with industrial Ph. D students, will form the Pro. Opt student group that will meet regularly with other researchers in Pro. Opt.

Pro. Opt Ph. D project 3: Cost, manufacturability and game theory in multi-physics optimization • This Ph. D project will have two central themes: – Cost and manufacturability in early design phases – Multi-objective optimization involving • • Multi-physics Multi-model Multi-level Modelling the Pro. Opt process (collaborative optimization? ) • The first theme means extending existing methods for topology optimization. • The second theme involves investigating use of game theory (Nash and Stackelberg games).

Pro. Opt • Ph. D project 3: Cost, manufacturability and game theory in multi-physics optimization Life Cycle Costs generally consist of – – Research and development Manufacturing • • • – – • Material Joining parts Features (e. g. drilling of a hole) Operating cost Disposal As a first step, cost of material and joining could be included in a tool for preliminary design

Pro. Opt Ph. D project 3: Multi-physics optimization • In a majority of industrial applications several physical domains are involved: this is obvious in high temperature structural applications (temperature and strength) or in heat exchangers (temperature and flow); in MEMS-applications electric fields couple to mechanics and in combustion chemistry and flow will be coupled. • Usually several objectives appear in multi-physics optimization. Thus, multiobjective optimization becomes a must.

Pro. Opt Ph. D project 3: Multi-model optimization • As an example of a multi-model problem we consider designing a car for both stiffness (körbarhet) and crashworthiness. • Clearly two models are needed. One operating in large deformation inelastic deformation and one that is a small displacement model. • Since several objectives appear, multiobjective optimization becomes a must.

Pro. Opt Ph. D project 3 – Multi-level optimization • Change of design space can also generate a multiobjective optimization problem. • In the buss example to the left one likes to use a continuous selection of sizes on a global level but a discrete (integer) on a local level.

Pro. Opt Ph. D project 3 - Approaches to multi-objective optimization • Goal programming (one objective is optimized while other are constraints) • Pareto optimality (computationally costly and gives many designs) • Weighted sum (gives some Pareto points but not all) • Nash, Stackelberg and/or other equilibrium concepts

Pro. Opt Ph. D project 3 – Game equilibrium • When different design goals can be associated to different groups of design variables various game equilibrium becomes a useful concept for multi objective optimization • Such an approach has advantages compared to more traditional Pareto optimization: we get one solution not a whole front • Applications: – Aerodynamic and structural objectives (wing design) – Crashworthiness and stiffness

Pro. Opt

Pro. Opt Ph. D project 3: Nash equilibrium for wing design The coupling of rib optimal design with optimal aerodynamic shape design is a typical example of Nash or Stackelberg type coupling between two objectives. Solution presentad by Altair

Pro. Opt Modelling the Pro. Opt process • The different phases in the overall design strategy are typically related as in Nash or Stackelberg games • Topology optimization may be seen as a follower and an optimization that uses such a result as input is a leader

Pro. Opt References – Project III • Kazuhiro Saitou, Kazuhiro Izui, Shinji Nishiwaki, Panos Papalambros, A survey of optimization in mechanical product development, Transactions of the ASME, Vol 5, 2005, 214 -226 • K Gabtois, A J Morris, The multi-disciplinary design of a large-scale civil aircraft wing taking account of manufacturing costs, Struct Multidisc Optim, Vol 28, 2004, 31 -46 • Z Tang, J-A Désidéri, J Périaux, Multicriterion aerodynamic shape design optimization and inverse problems using control theory and Nash games, J Optim Theory Appl, Vol 135, 22007, 599 -622