GENERATIVE DESIGN: ADVANCED DESIGN OPTIMIZATION
GENERATIVE DESIGN: ADVANCED DESIGNOPTIMIZATION PROCESSES FOR AERONAUTICALAPPLICATIONSS. Bagassi*, F. Lucchi*, F. De Crescenzio*, F. Persiani**Industrial Engineering Department, University of BolognaKeywords: Design Methods, Aircraft Design, Generative Design, Additive Manufacturing, Materialsand ProcessesAbstractCurrent researches on aircraft design aim toreduce airplanes and components ing to the challenge of reducing fuelconsumption and operational costs.In this perspective novel materials andtechnologies are developed, but also advancesin design methods and tools. Generative Designis a novel approach to automatically optimizecomponent design. The design process has to bedesigned itself to achieve the optimal solution,in relation to design parameters, requirementsand limits.Which peculiar features justify considering thistechnique to be a substantial step forward withrespect to classical MDO? Could GenerativeDesign be only an important, but notparticularly differentiated approach for thedesign of (aerospace) structures and possiblysystems of a higher level? For example, whenthe design goal is to find the best configurationof a structure, does generative design lead tothe discovery of new concepts, or types ofstructures, or it is a particular application ofgenetic algorithms to topological optimization?This paper aims to contribute to give an answerto the previous questions. Specifically, thegenerative design approach is expected to beable to select between basic concepts and usethese as the basic instructions and ingredientsof a recipe for the design of a new system.By these considerations, in this paper, werevised the improvements brought byGenerative Design principles within thetraditional design procedure in aeronautics,considering Additive Manufacturing technology.1 IntroductionMore flights, more potential passengers, feweremissions and shorter travel time will markfuture and challenges of aviation. According tosuch a vision, aircraft design is a ns, airliners’ requirements and aircraftperformances. Design methods and tools have todevelop accordingly. Innovative technologiesand materials will support designers insearching new solutions and processes. In thisway, innovation in design methods changescommon paradigms: the cross fertilizationcoming from incorporating concepts derivedfrom many disciplines and their integration withadvanced manufacturing systems and materialsmay improve design results.Inspired by the natural design processes andpatterns of nature, Generative Design is a designmethod for capturing the designer’s intent,generating new solutions. Characterized bydata-drivencollaborativecloud-basedtechnology, it relies upon a highly automatedactivity. A set of parameters and rules(commands to the designer) is considered as theDNA of the design process; rules andparameters are considered as the “genes”. Theyare combined by evolutionary algorithms oreven “brute force” computations. Theintroduction of such procedures into the designprocess allows the development of novel designsolutions by modifying the rules that define afinal design, difficult or impossible to achievevia other methods. Grammar-based techniquesexploit the principle of database amplification,1
S.BAGASSI, F.LUCCHI, F.DE CRESCENZIO, F.PERSIANIthe identification of rules, generating complexforms and patterns from simple specifications.Generative design principles received aparticular attention in architecture; some gooddefinitions coming from that field are reportedhere [1]:Ø “Generative design is not about designingthe building – It’s about designing the systemthat builds a building.” – Lars Hesselgren;Ø “Generative Design Processes is about themodelling of initial conditions of an object(its “genetics”) instead of modelling the finalform.” – Paola Fontana;Ø “Generative design systems are aimed atcreating new design processes that producespatially novel yet efficient and buildabledesigns through exploitation of currentcomputing and manufacturing capabilities” –Kristina Shea.1.1 Generative Design within the Design forAdditive Manufacturing approachNowadays it is generally accepted that inAerospace and Industrial applications, we canachieve the main advantages from theintroduction of advanced design procedures, ifwe combine them with Additive Manufacturing(AM) processes and techniques.Therefore, we have to take into account also thecoupling of evolutionary algorithms withinnovative manufacturing processes, likeAdditive Manufacturing, and new materials.This combination introduces more degrees offreedom in the final design concept: forexample, the mixing of materials with differentproperties allows having different propertiesdistributed in different zones of the same part,leading to multi-functional concepts.The opportunities offered by AM are notrestricted to multi-functional concepts. Considerthat, over the last years, AM’s adoption hasincreased across industries, with the aerospaceindustry contributing about 10.2% of AM’sglobal revenues in 2012. AM provides theflexibility to create complex part geometriesthat are difficult to build using traditionalmanufacturing, such as internal cavities orlattice structures that help reduce parts’ weightwithout compromising their mechanicalperformance [2]. Furthermore, AM’s impact oneconomies of scale and scope make it a naturalfit for Aerospace, which is largely gearedtoward customized production. A syntheticoverview of current and potential applications isreported in Figure 1.Fig.1 – Current and future applications of AM inAerospace industriesThe new system, once manufactured thanks toAM, should satisfy to the functionalrequirements in an innovative and moreefficient way, targeting also a simpler designand a substantial cost reduction. Novelstructural materials and advanced AMtechniques make these technologies ready to beintroduced within the generative design processalso for safety critical context, such as theaeronautical.Even if the advantages from the introduction ofthe couple Generative Design and AdditiveManufacturing has been widely considered,only some case studies have been provided inarchitecture,stylingandbiomedicalapplications, but has not be explored enough inaerospace.2 Generative Design approach in productdevelopment and aeronautical industryTypical Aircraft design practices consider thedesign process spread into three main phases:the conceptual design, the preliminary designand the detailed design. The design solution isproposed throughout the typical diverging –converging process, in relation to designrequirements and limits. Multidisciplinaryoptimization processes are currently beingdeveloped to support designer in assessing theoptimal solution, in relation to all designfeatures and parameters.Generative Design is a novel form-findingprocess that takes into account structuralperformances,materialpropertiesand2
GENERATIVE DESIGN: ADVANCED DESIGN OPTIMIZATIONPROCESSES FOR AERONAUTICAL APPLICATIONSergonomic demand, throughout an automaticiterative holistic approach for componenttopology optimization [3]. The design approachis grammar based, and the design activity itselfhas to be designed, on the basis of evolutionaryalgorithms in order to meet the optimal scoringfor an objective fitness function [4].Evolutionary design approaches limits toperforming the numerical optimization on theset of design parameters [5].On the other side, Topology Optimizationintends to find an optimal structuralconfiguration within a given design domain forspecified objectives, constraints, loads andboundary conditions. In general, it relates to sizeor shape optimization issues [6], , withparticular regards to component weightreduction: optimized configurations are definedre-distributing the material in the designdomains with the prescribed loads and boundaryconditions.Generative Design includes both EvolutionaryDesign and Topological Optimization, but it isnot limited to that.After years of research and development, somespecialized Generative Design tools arebecoming part of CAD/CAE platforms and areentering the market [8].A company developed software to be integratedwith industry-leading Building InformationModel (BIM) and CAD platforms. Thissoftware includes “two classes of tools: the firstclass connects existing tools together to allowseamless execution of complex workflows, andthe second class captures design intent.”Within the typical diverging – convergingdesign process, it is introduced the generativedesign domain for concept selection: the tradeoff between component performances, weightsand costs is simplified and the multidisciplinaryoptimization approach is improved, see Fig.2.Fig.2 – Scheme of Generative Design role in the designactivityEven if in simpler case studies, the generativerepresentation derived from evolutionarydesign, where the Evolutionary Design Systemis able to explore the space of design topologies,advanced Generative Design System and Toolsshould include some key aspects both in thedesign features and in the representation of thesolutions: modularity, regularity and hierarchy.Modules are at the base of the designrepresentation and they are hierarchicallyformed and could be combined or reused todefine the final topology optimization [9].From a computer programming language pointof view, rule-based design-constructionprograms define the design-programming tool.Subprocedure-like elements are at the basis ofthe modularity feature; iterative loops and subprocedures enable the regularity role and a nestbase approach defines the hierarchy, to link allthe sub-procedures and loops, within anassembly procedure.The final achieved design is optimized inaccordance of the proposed requirements andlimits and consists in a “family of designs”according to different parameters that could beused as input to the defined rules or changingthe objective fitness function (Figure 3) [9].3
S.BAGASSI, F.LUCCHI, F.DE CRESCENZIO, F.PERSIANIFig.3 Generative design examples: mutations of theresults according to encoded design changes [9]In [10], authors consider the digital designprocess as the composition of two mainactivities: physics-based processes, to performthe simulation of complex natural phenomenon,followed by a progressive formation andmutation to allow the creation and dynamicevolution of the simulation components.Thanks to its ability to generate novel,unconventional and complex structures (Fig. 4),developers and supporters consider GenerativeDesign one of the most promising methods to beexplored for evolving new designs, and suitedfor aerospace applications, even if it is notenough explored.Fig.4 GD outcome using Additive Manufacturingperformances [8] in the automotive field: design of aload-bearing engine block3 Methods and tools for Generative DesignDriving the design process, the human designerhas a central role to create and define the mostsuitable design solution. In this frameworknovel CAD tools have been developed tosupport designer in his creative role, and theyhave to meet the following requirements: The signer dependency is less demandingand less disruption to work processes; Allowing designers to navigate through thewhole design space; Supporting chaotic and unstructured workprocesses; Structured as an auxiliary tool; Support and enable the emergence of thedesigner in order to stimulate creativity; The content of the detailed design phase ofthe design to be effective.CAD tools enabled for Generative Designpractices do not have a pre-structured workflow;furthermore, the design development process iscouple with the generation of new knowledgeabout the design problem at each step of theiterative procedure.Recently several “generative design softwaretools” have been released, provided either aspart of CAE (Computer Aided Engineering)Suite such as Autodesk and Dassault Systèmesproducts or as stand alone applications (namelyGenoform and nTopology Element) withdifferent levels of interoperability with theexisting CAD software.Those tools that are part of CAE Suites exploitthe FEM tools to provide the user with a supporttool in designing full stressed components thatmaximize the efficiency of the structuralfunction. The design space is drawn by the useras the envelope the component should fit in.Loads and constraints are applied in a similarmanner as in the FEM tools, and the targetoptimization function is selected (e.g. minimumweight, maximum stiffness). As a result of theoptimization algorithm applied a single solutionis provided to the user by means of a roughmodel that should be then manually refined inorder to obtain an acceptable solution.Those tools are powerful from a structuralengineering point of view, however they are notfully capturing the creative intent of thedesigner.On the other hand, the generative design standalone applications (those that are not part of aCAE Suite) are providing the designer with a setof alternative product configurations that can beevaluated by the designer against aesthetical,structural or functional criteria using the4
GENERATIVE DESIGN: ADVANCED DESIGN OPTIMIZATIONPROCESSES FOR AERONAUTICAL APPLICATIONSappropriate tool for each relevant criterion.Genoforming parametric CAD systems arebased on CAD parametric models, they areplugin applications that work acting on theparameters of CAD programs by means ofgenetic algorithm. Specific features allow theuser to control the level of creative explorationby a slider bar (Genoform).4 Generative Design principles in componentdevelopment: case studiesSome case studies are presented in literature toassess the design improvements from theintroduction of Generative Design principleswithin the typical product developmentprocedure. Generally speaking, the designoptimization activity has the main objective toprovideasignificantweightsaving;furthermore, nowadays, advanced componentsare designed to be directly manufactured by anadditive technique.Generative design process starts with a designerdefining a design area, connection points linkedby parameters. An example can be seen inFigure 5, where these parameters are defined fora motorcycle swing arm design [11].Optimization is based on desired erance, cost, and strength of the part and itsability to withstand specified forces.Fig.5 Generative design of a motorcycle swing arm:bounding space (on the top left), connections (on thetop right), limits (on the bottom left) and design output(on the bottom left) [11]Innovative tools for Generative Design supportthe designer to fit all defined constraints, and tomeet requirements. An efficiently highautomation level support all design tasks,including aesthetics features and the definitionof innovative and effective models to explore,reducing the design time and improving thedesign results.Within Autodesk Research framework, ProjectDreamcatcher allow designer to iterative definesome most suitable solutions, testing thestrength and removing unnecessary material ateach design step [11]. The designer role is alsoto choose the best one and to modify itaccording to some specific additionalrequirements.Airbus and Autodesk have developed one of themost effective examples of Generative Designin aeronautics [12]. Cabin partitions aredesigned to be manufactured by a 3D additivemanufacturing process. Custom algorithms havebeen developed to generate unconventionalstructures to mimics cellular shapes and bonegrowth. Stronger and lighter micro-lattice newbionic partition structures are designed. Eachmodel is 45 per cent (30 kg) lighter than currentdesigns. When applied to A320 cabin it resultsin a reduction of 465,000 metric tons of C02emissions per year. Laser Powder Bed AdditiveLayer Manufacturing (ALM) processes is thetechnology used to manufacture the novelbionic partitions. Scalmalloy is the highperformance aluminium powder used for the 3Dprinting process, thanks to its high strength andtough properties, combined with low weight ofaluminium alloys.Fig.6 Novel bionic partition for Airbus A 320 cabininteriors [12]: the goals are met by coupling generativedesign with an additive manufacturing technique andthe introduction of novel materialsThe Generative Design approach considers adefined base model; a macro and microgeometry optimization process are iteratively5
S.BAGASSI, F.LUCCHI, F.DE CRESCENZIO, F.PERSIANIcarried out. At a macro scale, the algorithmsketch lines to network many reference points,while at a micro scale a logic similar to bonegrowth is applied to support the highest strengthareas of the structure (Figure 7).Fig.7 Design approach of the algorithm developedwithin Autodesk to meet Airbus cabin interiors byGenerative Design [12]ConclusionsGenerative Design is a novel procedure tosupport designer in widely explore the designspace. It is not only a topology optimization,neither an evolutionary algorithm, but itcombines several optimization modules totopology definition within a CAD environment,according to design requirements, limits and thebounding space. The output is not only the mostsuitable solution, while it is a family of differentresults that the designer could properly selectand modify. The solution space is generallydeveloped considering freeform shapes: itwould not be possible to reach a better solutionby means of a traditional design approach.Furthermore, the selected shape is designed tobe manufacture by an Additive Manufacturingprocess.Even if some case studies and some tools havebeen developed, the potentials brought byGenerative Design principles are not yetexplored enough. Some examples have beenconsidered in small component design, but onlyfew cases and tests are proposed in aeronautics.Low weight and structure’s strength used to bethe main objectives of aircraft componentdesign. In this framework the development ofrobust design procedure that include nts both in components’ feature anddesign results and in design time reduction aswell as aircraft operational costs, since a hugereduction of structures’ weight is foreseen.References[1] what-is-generative-desing/[2] Hod Lipson, Frontiers in Additive Manufacturing:The Shape of Things to Come, The BRIDGE NATIONAL ACADEMY OF ENGINEERING,Carol R. Arenberg editions, Vol. 42, No. 1, Spring2012[3] Hemmerling Marco, Ulrich Nether, Generico A casestudy on performance-based design, SIGraDi 2014Proceedings of the 18th Conference of theIberoamerican Society of Digital Graphics – Uruguay- Montevideo 12,13,14 November 2014, pp. 126-129,ISBN: 978-9974-99-655-7[4] McCormack, JP, Dorin, A & Innocent, TC,Generative design: a paradigm for design research,in J Redmond, D Durling & A de Bono (eds),Futureground. vol. 2, Monash University, Faculty ofArt & Design, Caulfield East Vic Australia, DesignResearch Society International Conference, CaulfieldEast Vic Australia, 1 January, 2005, ISBN0975606050[5] David Greiner, Blas Galván, Jacques Périaux,Nicolas Gauger, Kyriakos Giannakoglou, GabrielWinter, Advances in Evolutionary and DeterministicMethods for Design, Optimization and Control inEngineering and Sciences, Computational Methodsin Applied Sciences, Volume 36, 2015, SBN: 978-3319-11540-5, DOI: 10.1007/978-3-319-11541-2[6] Zhu, J., Zhang, W. & Xia, L., Topology Optimizationin Aircraft and Aerospace Structures Design, 1831-015-9151-2[7] Ji-Hong Zhu, Xiao-Jun Gu, Wei-Hong Zhang, PierreBeckers, Structural design of aircraft skin stretchforming die using topology optimization, Journal ofComputational and Applied Mathematics, Volume246, July 2013, Pages 278-288, ISSN 001[8] ] Gregory S. Hornby, Generative representation forComputer-Automated Evolutionary Design, NASAAmes Research Center[10] Ramtin Attar, Robert Aish, Jos Stam, DuncanBrinsmead, Alex Tessier, Michael Glueck, AzamKhan, Physics-based generative design, CAADFutures Conference, pp. 231-244, 2009.[11] garm[12] oneering-bionic-3D-printing.html6
GENERATIVE DESIGN: ADVANCED DESIGN OPTIMIZATIONPROCESSES FOR AERONAUTICAL APPLICATIONSContact Author Email AddressFurther information can be requested to thecorresponding author: f.lucchi@unibo.itCopyright StatementThe authors confirm that they, and/or their company ororganization, hold copyright on all of the original materialincluded in this paper. The authors also confirm that theyhave obtained permission, from the copyright holder ofany third party material included in this paper, to publishit as part of their paper. The authors confirm that theygive permission, or have obtained permission from thecopyright holder of this paper, for the publication anddistribution of this paper as part of the ICAS proceedingsor as individual offprints from the proceedings.7
the discovery of new concepts, or types of structures, or it is a particular application of genetic algorithms to topological optimization? This paper aims to contribute to give an answer to the previous questions. Specifically, the generative design approach is expected to be able
Combining information theoretic kernels with generative embeddings . images, sequences) use generative models in a standard Bayesian framework. To exploit the state-of-the-art performance of discriminative learning, while also taking advantage of generative models of the data, generative
1 Generative vs Discriminative Generally, there are two wide classes of Machine Learning models: Generative Models and Discriminative Models. Discriminative models aim to come up with a \good separator". Generative Models aim to estimate densities to the training data. Generative Models ass
Generative Design in Revit -Save Default Settings Generative Design in Revit -Drop-down Inputs Generative Design in Revit -Constant and Variable Inputs Generative Design in Revit -Manage Study Type Folders Dynamo for Revit 2.10 Multiple Routes for Path of Travel Spatial Grids for Documenting Layouts Autodesk Revit 2022
veloped for both generative and discriminative models. A straightforward, generative semi-supervised method is the expectation maximization (EM) algorithm. The EM ap-proach for naive Bayes text classification models is discussed by Nigam et al. in [17]. Generative semi-supervised meth-od
probabilistic generative models, which includes autoencoders[10] and powerful variants[13, 1, 14]. The second class, which is the focus of this paper, is called Generative Adversarial Networks (GANs)[5]. These networks combine a generative n
ple generative models based on different feature detectors and different numbers of parts) into a single classifier. Section 8 discusses the main results and observa-tions of this work. 2 Generative Models In this section we briefly review a class of generative models which will be used in conjunction with
designers design that are vitally important for understand-ing how to embed generative systems into human design processes. In Section 4 we point out some problems in con-ceptual design that interactive generative systems can alle-viate. Sections 5 and 6 present two examples of interactive automatic design systems based on psychological research
Artificial intelligence, or the idea that computer systems can perform functions typically associated with the human mind, has gone from futuristic speculation to present-day reality. When the AlphaGo computer program defeated Lee Sedol, a nine-dan professional master, at the game of Go in 2016, it signaled to the world that it is indeed possible for machines to think a bit like humans—and .
