SlideShare une entreprise Scribd logo
1  sur  179
Télécharger pour lire hors ligne
Master Thesis
                     (Dissertação de Mestrado)

       Master in Design and Product Development Engineering
(Mestrado em Engenharia da Concepção e Desenvolvimento de Produto)



                  Optimal Forms
  Generative Modeling Techniques in Optimization




                Nelson de Jesus Silvério da Silva




                           Leiria, July 2011
Master Thesis
                           (Dissertação de Mestrado)

        Master in Design and Product Development Engineering
 (Mestrado em Engenharia da Concepção e Desenvolvimento de Produto)



                        Optimal Forms
    Generative Modeling Techniques in Optimization




                      Nelson de Jesus Silvério da Silva

                        Scientific Adviser: Dr. Nuno Alves
(Escola Superior de Tecnologia e Gestão do Instituto Politécnico de Leiria, Portugal)

                       Scientific Co-Adviser: Dr. Eva Eggeling
                                (Fraunhofer, Austria)


                                  Leiria, July 2011
Report submitted to the Polytechnic Institute of Leiria in partial fulfillment of the
requirements for the degree of Master in Design and Product Development
Engineering (Mestrado em Engenharia da Concepção e Desenvolvimento de
Produto).




                         ISBN:
                         © Polytechnic Institute of Leiria




                                                                                    i
ii
The Jury




President




Vogals




            iii
iv
To My Family




           v
vi
Acknowledgments


    To my Scientific Advisor, Prof. Dr. Nuno Alves (Vice-director of CDRSP
    Research Centre), for the constant attention, motivation and ideas given
    throughout the course of the thesis and for helping me in achieving success.
    Thank you also for your friendship and whole-hearted smile.

    To my Scientific Co-Adviser, Dr. Eva Eggeling (Business Unit Manager of the
    Visual Computing Fraunhofer Austria Research GmbH), thank you so much
    for the support and incentive given to me personally, with your warm
    friendship and also for the support given professionally that made the
    realization of this thesis possible.

    To M.Sc. Torsten Ullrich (Researcher at Fraunhofer, Austria), for priceless
    knowledge, contribution, helps and of course, he‟s sincere friendship, this was
    crucial for the development of this innovative work.

    To Prof. Dr. Paulo Bártolo (Director of CDRSP Research Centre) and Prof. Dr.
    Helena Bártolo (CDRSP Research Centre) for all the support throughout the
    course of the thesis and for receiving me at the CDRSP Research Centre.

    To Prof. Dr. Dieter Fellner (Director of the Fraunhofer Institute for Computer
    Graphics Research - IGD) for giving me the opportunity to develop this work
    with the CGV Group and Fraunhofer Austria and for allowing me to have
    access to all the materials and valuable information produced within the
    CGV/Fraunhofer group.

    To M.Sc. Volker Settgast (Researcher at Fraunhofer Austria) for all the great
    tips about Autodesk Maya and Rendering in general and he‟s invaluable
    friendship.

    To Dr. Christina Lemke (Architect with projects in Germany and Spain, urban
    planner, construction biologist & ecologist) for her support, that kindly allowed
    me to have access to her published PhD thesis work.

    To my wife, for her patience, for being always there, supporting and
    encouraging and for understanding all the time I wasn‟t around.
    To ALL of you, that made this thesis possible, with your comprehension,
    motivation, support and encouragement.

    This master thesis was only possible to achieve, due to the good relationships
    and close partnership that was settled between CDRSP and Fraunhofer
    Austria. Thank you all so much…



                                                                                   vii
viii
Keywords   Procedural, Optimization, Evolutionary, Algorithm, Simulation,
           Building


Abstract   The generative modeling paradigm is a shift from static models to
           flexible models. A generative model describes a modeling process
           using functions, methods and operators. The result is an
           algorithmic description of the construction process. Each
           evaluation of such an algorithm creates a model instance, which
           depends on its input parameters (width, height, radius, orientation,
           etc.). These values are normally chosen according to aesthetic
           aspects and style. In this study, the model‟s parameters are
           automatically generated according to an objective function. A
           generative model can be optimized according to its parameters, in
           this way, the best solution for a constrained problem is determined.
           The field of application is energy and architecture. Besides the
           establishment of an overall framework design, this work consists
           on the identification of different building shapes and their main
           parameters, the creation of an algorithmic description for these
           main shapes and the formulation of the objective function,
           respecting a building‟s energy consumption (solar energy, heating
           and insulation). Also, this work aims the conception of an
           optimization pipeline, combining an energy calculation tool with a
           geometric scripting engine. In this study, one can read about state
           of the art developments related to architecture and procedural
           modeling. The major contribution of this development is to present
           methods that lead to an automated and optimized 3D shape
           generation for the projected building (based on the desired
           conditions and according to specific constrains), this will help in
           the construction of real buildings that account for less energy
           consumption and for a more sustainable world.




                                                                                  ix
x
Table of
Contents




           xi
xii
Acknowledgments ......................................................................................................................................... vii


      Keywords ........................................................................................................................................................ ix


      Abstract.......................................................................................................................................................... ix


Table of Contents....................................................................................................................................... xi


Figures List ...............................................................................................................................................xvii


Tables List ................................................................................................................................................ xxv


Abbreviations and Acronyms .................................................................................................................. xxix


I.       Introduction ......................................................................................................................................... 1

         1      The Problematic .................................................................................................................................. 1
         1.1       Thesis Structure ............................................................................................................................... 4


II.          State of the Art ................................................................................................................................. 7

         2      Overview ............................................................................................................................................. 7
         2.1       Parametric and Procedural Modeling .............................................................................................. 8
         2.1.1         Plugins for existent 3D Software: Blender, 3D Studio Max and Others ....................................... 9
         2.1.2         CityEngine .................................................................................................................................. 10
         2.1.3         Bentley – MicroStation Extension: GenerativeComponents (GC).............................................. 13
         2.1.4         Rhinoceros and Grasshopper ..................................................................................................... 14
         2.1.5         Generative Modeling Language (GML) ...................................................................................... 16
         2.1.6         Euclides Framework and JavaScript ........................................................................................... 18
         2.1.7         Autodesk Revit Architecture 2012 ............................................................................................. 19
         2.1.8         Project Vasari and Project Nucleus ............................................................................................ 21
         2.1.9         Autodesk Adaptive Components ............................................................................................... 23
         2.2       Evolutionary Architecture and the Use of Algorithms in Optimization of Problems ..................... 25
         2.2.1         Differential Evolution (DE) ......................................................................................................... 28
         2.2.2         Pros and Cons of using Evolutionary Algorithms (EAs) .............................................................. 38
         2.2.3         Other Evolutionary Based Algorithms - DE/EDA and Hybrid-DE ................................................ 39
         2.3       Advanced Rendering, Visualization and Interaction Techniques in Architecture ......................... 40
         2.3.1         Multitouch (MTT) ....................................................................................................................... 40
         2.3.2         Virtual and Augmented Reality .................................................................................................. 43
         2.3.3         Computer Generated Holography ............................................................................................. 47
         2.3.4         Advanced Rendering .................................................................................................................. 50



                                                                                                                                                                       xiii
2.3.5       Virtual World Interactivity ......................................................................................................... 53
       2.4      “A World Full of Sensors” .............................................................................................................. 57
       2.4.1       Sensors Feed Information into Virtual Worlds .......................................................................... 57
       2.4.2       Remote Monitoring of Persons Inside Buildings ........................................................................ 60
       2.4.3       Kinetic, Responsive Performative and Adaptive Architecture ................................................... 61
       2.5      Reverse Engineering and Rapid Prototyping ................................................................................. 62
       2.5.1       Reverse Engineering................................................................................................................... 63
       2.5.2       Rapid Prototyping ...................................................................................................................... 68
       2.5.3       Rapid Prototyping Techniques ................................................................................................... 69
       2.5.4       Personal Fabrication and Future Manufacturing ....................................................................... 76
       2.6      Building Information Modeling (BIM) and Automated Construction of Buildings ........................ 78
       2.6.1       Building Information Modeling (BIM) ........................................................................................ 78
       2.6.2       Automated Construction of Buildings ........................................................................................ 79


III.       Simulation Tools in Architecture ..................................................................................................... 81

       3     Outline............................................................................................................................................... 81
       3.1      EnergyPlus and DesignBuilder ....................................................................................................... 82
       3.2      Autodesk Ecotect Analysis ............................................................................................................. 84
       3.2.1       Short Comparison between Autodesk Ecotect and EnergyPlus ................................................. 87
       3.3      Ansys: AirFlow ............................................................................................................................... 88
       3.4      Sustainability Tools in Architecture – Comparison Studies/Audits ............................................... 90


IV.        A Global Optimization Framework ................................................................................................. 91

       4     Problematic of “Form Follows Energy” and the Pursuit of Solutions................................................ 91
       4.1      Answer to the Problematic: Optimal Forms - A Global Optimization Framework ........................ 97
       4.2      Identification of Essential Forms Used in the Real World ............................................................. 98
       4.2.1       Procedural Shape Generation .................................................................................................... 99
       4.2.2       Code Writing Using Euclides and JavaScript ............................................................................ 101
       4.3      Simulation Tools Integration ....................................................................................................... 102
       4.3.1       Simulation in Ecotect and the Admittance Method................................................................. 104
       4.3.2       Initial Manual Workflow Tests ................................................................................................. 105
       4.4      Differential Evolution................................................................................................................... 107
       4.5      The Developed Global Optimization Framework ........................................................................ 110
       4.6      Case Studies and Presentation of Results.................................................................................... 112
       4.6.1       Case Study 1 – Classic Shape Building Optimization ................................................................ 113
       4.6.2       Case Study 1 - Presentation of Results..................................................................................... 116
       4.6.3       Case Study 2 – Cube Shape Building Optimization .................................................................. 126




xiv
V.        Conclusions and Future Work ....................................................................................................... 131

      5     Summary ......................................................................................................................................... 131
      5.1      Final Conclusions ......................................................................................................................... 132
      5.2      Future Work................................................................................................................................. 133


REFERENCES ............................................................................................................................................ 135




                                                                                                                                                              xv
xvi
Figures List




               xvii
xviii
FIG. II-1 – SUICIDATOR CITY GENERATOR (SCG) FOR BLENDER                                                              9
FIG. II-2 – PROCEDURAL/PARAMETRIC EXAMPLES CREATED IN CITYENGINE                                                    10
FIG. II-3 – CITYENGINE IDE AND THE RULE EDITOR CAPABILITIES                                                         12
FIG. II-4 – GENERATIVECOMPONENTS (GC) IDE, BENTLEY MICROSTATION                                                     13
FIG. II-5 – RHINOCEROS IS USED IN MULTIPLE FIELDS, INCLUDING ARCHITECTURE                                           14
FIG. II-6 – VORONOI EXAMPLES, CREATED USING RHINOCEROS AND GRASSHOPPER, BY ATSUO NAKAJIMA (TOKYO, JAPAN)            15
FIG. II-7 - PARAMETRIC STRATEGIES ACHIEVED USING RHINO AND GRASSHOPPER. (CREATED BY THE AUTHOR OF THIS MASTER
      THESIS)                                                                                                       15
FIG. II-8 - CREATION OF A SIMPLE HOUSE MODEL USING GML, THE EXTRUDE OPERATOR IS REPEATEDLY APPLIED TO THE GROUND
      POLYGON. TO CREATE THE ROOF, THE COMBINED OPERATOR COLLAPSE-MID IS APPLIED TO THE FACECW AND FACECCW

      EDGES OF THE EDGE RETURNED BY THE EXTRUDE OPERATION.                                                          16
FIG. II-9 – PARAMETERIZATION/CONFIGURATION OF A CHAIR WITH GML                                                      17
FIG. II-10 – GOTHIC STYLE BUILDING GENERATED WITH GML                                                               17
FIG. II-11 – CONFIGURATION OF DIFFERENT WHEEL RIM STYLES USING GML                                                  17
FIG. II-12 – EXAMPLE OF A 3D APPLICATION CREATED FOR THIS THESIS USING EUCLIDES AND JAVASCRIPT. IT ALLOWS THE
      CONTROL OF SEVERAL SHAPE PARAMETERS ON THE “CLASSIC BUILDING FORM EXAMPLE”.                                   18
FIG. II-13 – ENERGY CONSUMPTION STUDY USING AUTODESK REVIT                                                          19
FIG. II-14 – CONCEPTUAL DESIGN IN AUTODESK REVIT ARCHITECTURE                                                       20
FIG. II-15 – SUN STUDIES USING PROJECT “VASARI”                                                                     21
FIG. II-16 - PANEL STUDY USING REVIT AND VASARI                                                                     22
FIG. II-17 - USING REVIT, VASARI AND NUCLEUS PHYSICS FOR A PANEL STUDY, PLUS ANALYSIS                               22
FIG. II-18 – ADAPTIVE COMPONENTS IN AUTODESK REVIT                                                                  23
FIG. II-19 – ADAPTIVE PANEL EXAMPLE IN AUTODESK REVIT                                                               24
FIG. II-20 – ANOTHER ADAPTIVE PANEL EXAMPLE, BUILT USING ADAPTIVE COMPONENTS                                        24
FIG. II-21 – IMAGE TAKEN FROM THE BOOK “THE SELFISH GENE” BY RICHARD DAWKINS                                        25
FIG. II-22 – SEVERAL VISIONS RELATED TO EVOLUTIONARY ARCHITECTURE AND BIOMIMETIC                                    26
FIG. II-23 – EVOLUTIONARY EXAMPLES TAKEN FROM THE BOOK “AN EVOLUTIONARY ARCHITECTURE” BY JOHN FRASER [23]           26
FIG. II-24 – DYNAMIC GEOMETRY COMPUTATION, “SHANGHAI TOWER - GEOMETRY GENERATE AND RENDERING” (MICHAEL
      PENG)                                                                                                         27
FIG. II-25 – ECOLOGICAL HOUSE OF THE FUTURE (EUGENE TSUI)                                                           27
FIG. II-26 – THE GLOBAL OPTIMIZATION PROBLEM (EXAMPLE IN MATLAB). SEARCH OF THE HIGHEST MOUNTAIN PEAK AMONG A
      NEIGHBORHOOD OF OTHER HIGH MOUNTAINS PEAKS.                                                                   28
FIG. II-27 - OBSERVATION, ANALYSIS AND COMPUTATION OF BRANCHING PATTERNS IN NATURAL SYSTEMS (BY EVAN GREENBERG
      [29])                                                                                                         29
FIG. II-28 – SIMPLE EA’S STEPS                                                                                      32
FIG. II-29 – GENERAL EVOLUTIONARY ALGORITHM: I: INITIALIZATION, F(X): EVALUATION, ?: STOPPING CRITERION, SE: SELECTION,
      CR: CROSS-OVER, MU: MUTATION, RE: REPLACEMENT, X*: OPTIMUM. AUTHOR: JOHANN "NOJHAN" DRÉO                      33




                                                                                                                    xix
FIG. II-30 - DE OPTIMIZATION PERFORMANCE (PEDERSEN, M. [31]) ON SEVERAL DIFFERENT PROBLEMS USING DE/RAND/1/BIN
      ALGORITHM. PLOTS SHOW THE MEAN FITNESS ACHIEVED. OVER 50 OPTIMIZATION RUNS                                  34
FIG. II-31 – 3D ANAGLYPH AND ACTIVE STEREO VISUALIZATION ON A MULTITOUCH DI TABLE (MTT4ALL MULTITOUCH TABLE
      WAS BUILT BY THE AUTHOR OF THIS MASTER THESIS)                                                              41
FIG. II-32 – INITIAL MTT4ALL SCALE MODEL (LEFT) AND FINAL MTT4ALL FUNCTIONAL PROTOTYPE IN USE , RUNNING
      FRAUNHOFER VIRTUALDESK APPLICATION, DEVELOPED BY THE AUTHOR OF THIS MASTER THESIS FOR FRAUNHOFER AUSTRIA
      (RIGHT)                                                                                                     42
FIG. II-33 – FRAUNHOFER, MULTITOUCH ARCHITECTURE VISUALIZATION – MESSE FRANKFURT GMBH (USING THE
      INSTANTREALITY FRAMEWORK)                                                                                   43
FIG. II-34 – ANAGLYPH VISUALIZATION (CREATED BY THE AUTHOR OF THIS MASTER THESIS)                                 44
FIG. II-35 – ACTIVE STEREO VISUALIZATION (CREATED BY THE AUTHOR OF THIS MASTER THESIS)                            44
FIG. II-36 – AN AUGMENTED REALITY SYSTEM DEVELOPED BY FRAUNHOFER (MONITOR + CAMERA + VIRTUAL REALITY
      SOFTWARE)                                                                                                   45
FIG. II-37 – DAVE, CGV AUSTRIA: IMAGES ARE PROJECTED ON THE BACK PROJECTION SIDE WALLS AND ON THE FLOOR FROM
      ABOVE, MMIRRORS ARE USED TO REDUCE THE SPACE NEEDED                                                         45
FIG. II-38 – CGV AUSTRIA, THE DAVE, A 3D IMMERSIVE SYSTEM (EXPLORING THE NATIONAL LIBRARY OF VIENNA)              46
FIG. II-39 – THE SWEETHOME3D (MODELING APPLICATION) OUTPUT IS TRANSFERRED WITH A WEB SERVICE TO AN OPENSG
      CAVE APPLICATION WHICH LETS THE USER WALK THROUGH A 3D REPRESENTATION OF THE HOUSE PLAN                     46
FIG. II-40 – HEYEWALL (HIGH RESOLUTION MULTITOUCH SCREEN)                                                         47
FIG. II-41 – COMPUTER GENERATED HOLOGRAPHY. A COMPUTER CALCULATES A HOLOGRAPHIC FRINGE PATTERN FOR DISPLAY BY
      THE SPATIAL LIGHT MODULATOR (SLM), WHICH DIFFRACTS LASER LIGHT TO YIELD AN INTERACTIVE, TRUE 3D IMAGE       48
FIG. II-42 – TRADESHOW (AN HOLOGRAM SYSTEM)                                                                       49
FIG. II-43 – ALTHOUGH HE WAS IN MELBOURNE, TELSTRA'S CHIEF TECHNOLOGY OFFICER, HUGH BRADLOW (RIGHT), MAKES IS
      PRESENCE FELT AT A CONFERENCE IN ADELAIDE (PHOTO: TELSTRA)                                                  49
FIG. II-44 – TOUCHABLE HOLOGRAPHY INTERACTION SYSTEM. AN AERIAL IMAGING SYSTEM, A NON-CONTACT TACTILE DISPLAY
      AND A WIIMOTE-BASED HAND-TRACKING SYSTEM ARE COMBINED. IN THIS FIGURE, THE ULTRASOUND IS RADIATED FROM

      ABOVE AND THE USER FEELS AS IF A RAIN DROP HITS HIS PALM                                                    50
FIG. II-45 – 3D PHOTOREALISTIC RENDERING (CREATED BY HARCHI, AN ARCHITECTURE COMPANY BASED IN PORTUGAL)           51
FIG. II-46 – IPL/CDRSP FUTURE BUILDING (RENDERED IN AUTODESK MAYA 2011 BY THE AUTHOR OF THIS MASTER THESIS) 52
FIG. II-47 – MEDIUM QUALITY RENDERING OF A FACTORY INSTALLATION (CREATED IN DEEP EXPLORATION BY THE AUTHOR OF
      THIS THESIS)                                                                                                52
FIG. II-48 – HIGH QUALITY REAL-TIME INTERACTIVE RENDERING (WWW.ICREATE3D.COM)                                     53
FIG. II-49 – MANIPULATION OF VR DATA, PROVIDED BY MEMPHIS [1] (TICIVIEW VR SYSTEM). IGI/FRAUNHOFER RESEARCH
      GROUP, 2007, SEOUL - SOUTH KOREA, (THE AUTHOR OF THIS MASTER THESIS WAS A MEMBER IN THE TEAM RESPONSIBLE

      FOR THE DEVELOPMENT OF THIS SYSTEM)                                                                         55
FIG. II-50 – ADVANCED INTERACTIVE VISUALIZATION (USING IVIEWER), STARTING FROM LEFT TO RIGHT: (A) SKYSCRAPPER; (B) AN
      APARTMENT INSIDE THE SKYSCRAPPER (WWW.ICREATE3D.COM)                                                        55




xx
FIG. II-51 – VIRTUAL FACTORY SIMULATION/SERIOUS GAME THAT WILL ALLOW A COMPANY TO GIVE TRAINING TO USERS, (THIS
      PROJECT WAS CREATED AT CDRSP BY THE AUTHOR OF THIS MASTER THESIS)                                           56
FIG. II-52 – SCREENSHOT OF THE TRICORDER DEVICE SHOWING THE FLOORPLAN OF A LAB, OVERLAYED WITH PLUG ICONS TO
      REPRESENT SOUND, LIGHT, CURRENT CONSUMPTION, MOTION AND VIBRATION. ALSO AVERAGE DATE FROM ALL SENSORS IS

      DISPLAYED [52]                                                                                              58
FIG. II-53 – A PORTAL IN SECOND LIFE SHOWS SENSOR DATA OVER TIME                                                  59
FIG. II-54 – A VIRTUAL DATAPOND IN THE VIRTUAL ATRIUM (LEFT) AND A REAL DATAPOND IN THE REAL MEDIA LAB ATRIUM
      (RIGHT)                                                                                                     60
FIG. II-55 – DIFFERENT REPRESENTATIONS OF A PERSON DETECTION IN 3D, STARTING FROM LEFT TO RIGHT: (A) BILLBOARD WITH
      A THIN COLORED SURROUNDING LINE, (B) ADDITIONAL GEOMETRY, (C, D) OVERLAY MARKER WHICH IS NOT OCCLUDED BY

      THE SCENE                                                                                                   60
FIG. II-56 – A MODEL WHICH IS A KINETIC PAVILION THAT REACTS ON WEATHER DATA                                      61
FIG. II-57 – PHASES OF THE REVERSE ENGINEERING PROCESS                                                            63
FIG. II-58 – REVERSE ENGINEERING - CLASSIFICATION TECHNIQUES FOR 3D DATA ACQUISITION                              64
FIG. II-59 – STEINBICHLER COMET 5 PHOTOGRAMMETRY 3D SCAN EQUIPMENT AT CDRSP REVERSE ENGINEERING LABORATORY
                                                                                                                  65
FIG. II-60 – A 3D SCAN OF A REAL GRAPHITE ELECTRODE USED IN MOLDS INDUSTRY. (3D SCAN AND ANALYSIS/INSPECTION DONE
      BY THE AUTHOR OF THIS MASTER THESIS, USING STEINBICHLER COMET 5 EQUIPMENT, AND STEINBICHLER COMET PLUS AND

      COMET INSPECT SOFTWARE)                                                                                     65
FIG. II-61 – “SANTUÁRIO DO SENHOR DA PEDRA”, ÓBIDOS, PORTUGAL                                                     66
FIG. II-62 – POINTS OBTAINED FOR THE “SANTUÁRIO DO SENHOR DA PEDRA” BUILDING                                      66
FIG. II-63 – 3D CLOUD OF POINTS FOR THE “SANTUÁRIO DO SENHOR DA PEDRA” BUILDING                                   66
FIG. II-64 - A) STL AND B) 3D MODEL                                                                               67
FIG. II-65 – SCALE MODEL OF “SANTUÁRIO DO SENHOR DA PEDRA” OBTAINED BY RAPID PROTOTYPING                          67
FIG. II-66 – MAIN COMPONENTS OF A STEREO-LITHOGRAPHY MACHINE                                                      69
FIG. II-67 – PROTOTYPES PRODUCED USING STEREO-LITHOGRAPHY                                                         70
FIG. II-68 – SIMPLIFIED FDM PROCESS                                                                               71
FIG. II-69 – SCALE MODELS OBTAINED USING FDM                                                                      72
FIG. II-70 – SCHEME OF THE LOM PROCESS                                                                            73
FIG. II-71 – SCALE MODEL FOR OPORTO MUSIC HOUSE, PRODUCED USING LOM (PORTUGAL)                                    73
FIG. II-72 – SCHEME OF THE SELECTIVE LASER SINTERING (SLS)                                                        74
FIG. II-73 - 3D PRINTING PROCESS                                                                                  75
FIG. II-74 – COMPLETE SCALE MODEL OBTAINED USING THE 3D PRINTING PROCESS                                          75
FIG. II-75 – COMBINING SEVERAL TECHNOLOGIES/PROCESSES                                                             76
FIG. II-76 - A BRUSH MADE IN A 3D PRINTER, USING TWO DIFFERENT MATERIALS, PRINTED SIMULTANEOUSLY INTO A SINGLE AND
      NOT ASSEMBLED FUNCTIONAL OBJECT (OBJECT INC.)                                                               77




                                                                                                                  xxi
FIG. II-77 – BIM VIRTUAL INFORMATION (VIRTUAL SIMULTANEOUS VISUALIZATION OF SIX DIFFERENT PHASES OF AN ONGOING
       BUILDING PROJECT, CREATED USING GRAPHISOFT ARCHICAD PLATFORM)                                                  79
FIG. II-78 – VISION FOR AN AUTOMATED SYSTEM FOR AUTONOMOUS CONSTRUCTION OF BUILDINGS (BEHROKH KHOSHNEVIS) 80
FIG. II-79 – A REAL PROTOTYPE FOR AN AUTOMATED SYSTEM THAT WILL ALLOW THE CREATION OF BUILDINGS                       80
FIG. III-1 – ENERGYPLUS SIMULATION ZONES                                                                              82
FIG. III-2 – DESIGNBUILDER AND ITS BUILDINGS ENERGY EFFICIENCY RATING                                                 83
FIG. III-3 - WORKING IN ENERGYPLUS-MODE INSIDE ECOTECT, WHEN DEFINING OPERATIONAL SCHEDULES                           84
FIG. III-4 - INTERNAL DAYLIGHT FACTORS SHOWN OVER A STANDARD WORKING PLANE                                            85
FIG. III-5 - OVERLAYING A SUN-PATH ON THE MODEL VIEW                                                                  85
FIG. III-6 - ANNUAL CUMULATIVE SOLAR RADIATION OVER THE EXTERNAL SURFACES                                             86
FIG. III-7 – COLOURED CONTOURS OF THERMAL CONFORT IN A CONFERENCE ROOM PREDICTED FOR A PARTICULAR VENTILATION
       SYSTEM DESIGN (ANSYS, INC. PROPRIETARY)                                                                        88
FIG. III-8 – ANSYS CFD MODELLING OF REGIONAL FLOW PATTERNS NEAR CAPE SHOPPING CENTRE (STEPHAN SCHMITT &
       THOMAS KINGSLEY; QFINSOFT, SA)                                                                                 89
                                                                3
FIG. IV-1 – DIFFERENT POSSIBLE BUILDINGS FORMS, ALL WITH 1000M OF VOLUME, THIS CAN ALLOW THE COMPARISON OF
       RESULTS OBTAINED WITH DIFFERENT FORMS (PHD THESIS OF CHRISTINA LEMKE [88])                                     92
FIG. IV-2 – DEFINITION OF AN ELEMENTARY VOLUME, ACCORDING TO DEPECKER, P., ET AL. [91]                                93
FIG. IV-3 – FLOW FIELD AT A STREET INTERSECTION WITH A TALL BUILDING, ILLUSTRATING EXCHANGES BETWEEN THE STREETS AND
       ADDITIONAL MIXING PROCESSES DUE TO THE LARGE BUILDING                                                          94
FIG. IV-4 – VIEW OF GREENHOUSE SHAPES IN E-W ORIENTATION                                                              96
FIG. IV-5 – 3D MODELS THAT REPRESENT REAL WORLD FACTORIES                                                             98
FIG. IV-6 - SELECTED BASIC FORMS INSPIRED IN REAL WORLD BUILDING SHAPES, STARTING FROM LEFT TO RIGHT: (A) CUBE, (B)
       CLASSIC AND (C) CYLINDER) CREATED USING EUCLIDES AND RENDERED USING DEEP EXPLORATION                           98
FIG. IV-7 – TESTS FOR CREATING DIFFERENT 3D SHAPES (AND CONTROLLING ITS PARAMETERS) USING PARAMETRIC EQUATIONS 99
FIG. IV-8 – DEFINITION OF PARAMETRIC EQUATIONS IN MAPPLE 14                                                       100
FIG. IV-9 - OPTIMIZED CODE GENERATION PRODUCED BY MAPPLE 14                                                       100
FIG. IV-10 – PIECE OF JAVASCRIPT CODE TO GENERATE THE 3D CYLINDER SHAPE (CODE ADAPTED FROM PARAMETRIC EQUATIONS
       AND MAPPLE 14)                                                                                             101
FIG. IV-11 – RESULTING JAVASCRIPT/EUCLIDES INTERFACE THAT ALLOWS THE CONTROL OF EACH SHAPE PARAMETER              101
                                                       3
FIG. IV-12 - 3D SHAPE GENERATION IN EUCLIDES (1.000 M OF VOLUME); FOLLOWED BY THERMAL ANALYSIS; AND
       PRESENTATION OF THE ANALYZED SHAPE. OTHER TYPES OF ANALYSIS CAN BE PERFORMED AS WELL.                      105
                                                                             3
FIG. IV-13 – ANOTHER 3D SHAPE GENERATION IN EUCLIDES (MAINTAINING 1.000 M ); FOLLOWED BY THERMAL ANALYSIS; AND
       PRESENTATION OF THE ANALYZED SHAPE                                                                         106
FIG. IV-14 – EXAMPLE LIST FOR MATERIALS THAT CAN BE USED INSIDE AUTODESK ECOTECT                                  109
FIG. IV-15 - OVERVIEW OF “OPTIMAL FORMS”, THE DEVELOPED AND PROPOSED GLOBAL OPTIMIZATION FRAMEWORK                110




xxii
FIG. IV-16 – THE GLOBAL OPTIMIZATION FRAMEWORK RUNNING AUTONOMOUSLY (SIMULATION IN ECOTECT FOLLOWED BY 3D
      SHAPE GENERATION IN ORDER TO EVOLVE A POPULATION OF NEW BUILDINGS WITH DIFFERENT PARAMETERS USING THE

      DIFFERENTIAL EVOLUTION ALGORITHM)                                                                            111
FIG. IV-17 – THE ALGORITHM CHOOSES DIFFERENT INDIVIDUALS (BUILDINGS) FOR ANALYSIS, NOT STOPPING ON THE LOCAL
      MINIMA THAT WAS FOUND ALONG THE OPTIMIZATION PROCESS AND GIVING ROOM FOR JUMPING THOSE SAME LOCAL

      MINIMA                                                                                                       116
FIG. IV-18 – IN THIS RUN IT WAS GENERATED A COMPLETELY DIFFERENT INDIVIDUAL (BUILDING SHAPE), BUT THE ADMITTANCE
      VALUE WAS REALLY HIGH AND OTHER BETTER INDIVIDUALS WERE FOUND                                                117
FIG. IV-19 – A NEW OPTIMIZATION RUN, THIS TIME USING A DIFFERENT VALUE FOR CROSSOVER (0.6)                         118
FIG. IV-20 – OTHER “TIGHTER” CONSTRAINS WERE CHOSEN, AND THE RESULTS WERE SLIGHTLY WORST THEN THE INITIAL
      ATTEMPTS (RUNS 1, 2 AND 3) WHERE A WIDER DOMAIN OF SEARCH WAS USED                                           119
FIG. IV-21 – THE OPTIMIZATION RUNS (1, 2 , 3 AND 4) ARE PLOTTED HERE SIMULTANEOUSLY                                120
FIG. IV-22 – BY ALLOWING THE ALGORITHM TO GENERATE BUILDINGS THAT COULD USE A DIFFERENT ORIENTATION (LESS
      CONSTRAINED REGARDING THE ORIENTATION OF THE BUILDING), MORE EVALUATIONS WERE NEEDED, BUT A MUCH BETTER

      RESULT WAS OBTAINED                                                                                          121
FIG. IV-23 – A NEW CONSECUTIVE OPTIMIZATION RUN (USING EXACTLY THE SAME VALUES) IN ORDER TO CHECK IF THE BEHAVIOR
      WAS CONSISTENT FROM RUN TO RUN                                                                               122
                                                                                                     O         O
FIG. IV-24 – BY ALLOWING THE ALGORITHM TO SEARCH FOR SOLUTIONS IN AN ORIENTATION DOMAIN BETWEEN 0 AND 360
      AND BECAUSE THE BUILDING DOES NOT HAVE DOORS OR WINDOWS YET, THE ALGORITHM FOUND A GOOD SOLUTION BY

      ORIENTING THE BUILDING ON A DIFFERENT DIRECTION (WHEN COMPARED TO RUNS 5 AND 5_1)                            123
FIG. IV-25 – BY PLOTTING ALL THE INDIVIDUALS GENERATED BY THE GLOBAL OPTIMIZATION FRAMEWORK FOR RUNS (5, 5_1 AND
      6) IT’S POSSIBLE TO CHECK THE CONSISTENCE OF RESULTS OBTAINED ON THESE MORE COMPLETE OPTIMIZATION RUNS 124
FIG. IV-26 – CASE STUDY 1 (FINAL CLASSIC BUILDING SHAPE) – DAILY AND ANNUAL SUN PATHS + SHADOWS AND DAYLIGHT
                                    TH
      LEVELS AT 14:00 PM, FOR THE 8 OF SEPTEMBER (BASED ON SPECIFIC CLIMATE/WEATHER FILES FOR COIMBRA,
                                                               O                      3                   O
      PORTUGAL); WIDTH = 20 M; HEIGHT = 6 M; ROOF ANGLE = 60 ; VOLUME = 1000 M ; ORIENTATION: 122,10               125
FIG. IV-27 – WE CAN OBSERVE THAT AT EVALUATION 289 THE OPTIMIZATION FRAMEWORK HAD ALREADY ACHIEVED A VERY GOOD
      RESULT (COMPARED TO THE FINAL RESULT), BUT BECAUSE THE STOP CRITERION USED WAS, 1000 EVALUATIONS OR DX =

      1.0, THE OPTIMIZATION CONTINUED UNTIL DX = 1.0 FOR FOUR CONSECUTIVE TIMES                                    128
FIG. IV-28 - CASE STUDY 2 (FINAL CUBE BUILDING SHAPE) – DAILY AND ANNUAL SUN PATHS + SHADOWS AND TOTAL
                                              TH
      RADIATION LEVELS AT 14:00 PM, FOR THE 8 OF SEPTEMBER (BASED ON SPECIFIC CLIMATE/WEATHER FILES FOR
                                                                          3                      O
      COIMBRA, PORTUGAL); WIDTH = 10 M; HEIGHT = 6 M; VOLUME = 1000 M ; ORIENTATION: 121,86                        129




                                                                                                                   xxiii
xxiv
Tables List




              xxv
xxvi
TABLE III-1 - CHARACTERISTICS OF TWO DIFFERENT SIMULATION TOOLS [83] ................................................................. 87
TABLE IV-1 – ACTIVITY LEVEL IN AUTODESK ECOTECT ............................................................................................... 108
TABLE IV-2 - THIS TABLE SHOWS SEVERAL RUNS USING THE DEVELOPED FRAMEWORK IN ORDER TO OPTIMIZE THE DESIGN OF A
        “CLASSIC SHAPE FORM” BUILDING USING THE DIFFERENTIAL EVOLUTION (DE) ALGORITHM. IN THIS TABLE IS ALSO POSSIBLE
        TO SEE THE DIFFERENT CONSTRAINS AND PARAMETERS USED IN THE OPTIMIZATION PROCESS .................................... 114

TABLE IV-3 - RESULTS OBTAINED IN EACH OF THE OPTIMIZATION RUNS (ACCORDING TO TABLE IV-2).................................. 115
TABLE IV-4 – THIS TABLE SHOWS AN OPTIMIZATION RUN USING THE DEVELOPED FRAMEWORK IN ORDER TO OPTIMIZE THE DESIGN
        OF A “CUBE SHAPE FORM” BUILDING USING THE DIFFERENTIAL EVOLUTION (DE) ALGORITHM. IN THIS TABLE IS ALSO

        POSSIBLE TO SEE THE DIFFERENT CONSTRAINS AND PARAMETERS USED IN THE OPTIMIZATION PROCESS ....................... 127

TABLE IV-5 - RESULTS OBTAINED IN EACH OF THE OPTIMIZATION RUNS (ACCORDING TO TABLE IV-4).................................. 128




                                                                                                                                             xxvii
xxviii
Abbreviations
and Acronyms




            xxix
xxx
A                                                         D

ANM: Annealed Nelder and Mead strategy · 37               DAVE: Definitely Affordable Virtual Environment
AR: Augmented Reality · 42, 44, 53, 54, 55                    (Immersive VR System Developed by Fraunhofer) ·
ASA: Adaptive Simulated Annealing · 37                        45, 46
ASHRAE: American Society of Heating, Refrigerating        DDE: Dynamic Data Exchange · 87
    and Air Conditioning Engineers · 87                   DE: Differential Evolution · 4, 28, 32, 33, 34, 35, 36, 37,
                                                              39, 107, 112, 114, 127

B
                                                          E
BGA: Breeder Genetic Algorithm · 37
BIM: Building Information Modeling · 11, 19, 21, 78, 79   EA: Evolutionary Algorithm · 28, 32, 38
BLAST: Building Loads Analysis and System                 EASY: Evolutionary Algorithm with Soft Genetic
    Thermodynamics · 82, 83                                   Operators · 37
                                                          EDA: Estimation of Distribution Algorithm · 39
                                                          ES: Evolutionary Strategies · 37
C
                                                          ESRI: Environmental Systems Research Institute · 10
                                                          ESTG: Superior School of Technology and Management
CAD: Computer Aided Design · 8, 10, 14, 61, 63, 65, 78,
                                                              ·2
    79, 125, 133
                                                          Euclides: Fraunhofer JavaScript Procedural Modeler ·
CAM: Computer Aided Manufacturing · 14
                                                              2, 4, 5, 18, 81, 97, 98, 100, 101, 102, 103, 105, 106,
CAVE: Cave Automatic Virtual Environment · 42, 45, 46
                                                              107, 109, 112
CC: Contour Crafting (Automated Construction System)
    · 80
CDRSP: Centre for Rapid and Sustainable Development       F
    of the Product · vii, 2, 7, 41, 43, 52, 56, 65, 68
CEP: Complex Event Processing · 59                        FAR: Floor Area Ratio · 12
CFD: Computational Fluid Dynamics · 88, 89, 133           FCT: Portuguese Foundation for Science and
CGA: Computer Generated Architecture (shape                   Technology · 7
    grammar) · 11                                         FDM: Fused Deposition Modeling · 68, 70, 71, 72
CGH: Computer Generated Holography · 47, 48, 49
CIBSE: Chartered Institution of Building Services
                                                          G
    Engineers · 87, 104, 105, 141
CNC: Computer Numeric Control · 14
                                                          GA: Genetic Algorithm · 30
CPU: Central Processing Unit · 51, 52
                                                          GC: Bentley Microstation GenerativeComponents · 13
CR: Crossover · 34
                                                          GFA: Gross Floor Area · 12
CSG: Constructive Solid Geometry · 16
                                                          GIS: Geographic Information Systems · 10, 40
CT: Computed Tomography · 51
                                                          GML: Generative Modeling Language · 16, 17, 18
                                                          GPU: Graphics Processing Unit · 51, 52




                                                                                                                 xxxi
H                                                           O

HDE: Hybrid Differential Evolution · 39                     OLED: Organic Light-Emitting Diode · 58
HSM: High Speed Machining · 68
HVAC: Heating, Ventilation and Air Conditioning · 82,
                                                            P
    83

                                                            PDM: Product Data Management · 8
I                                                           PLM: Product Lifecycle Management · 8


ICEO: IEEE Competition on Evolutionary Optimization ·
                                                            R
    37
IEEE: Institute of Electrical and Electronics Engineers ·
                                                            Rhino: (a.k.a. Rhinoceros), it's a commercial NURBS-
    37
                                                                based 3D modeling tool, developed by Robert
IPL: Polytechnic Institute of Leiria · 2, 52
                                                                McNeel & Associates · 14; Rhinoceros (Robert
                                                                McNeel & Associates) · 14, 15
L                                                           RICS: Royal Institution of Chartered Surveyors · 90, 140


LUA: Lightweight multi-paradigm programming
                                                            S
    language designed as a scripting language with
    extensible semantics as a primary goal · 87, 102,
                                                            SCG: Suicidator City Generator · 9
    103, 141
                                                            SDE: Stochastic Differential Equations · 37
                                                            SLM: Spatial Light Modulator · 48
M                                                           SLS: Selective Laser Sintering process · 68, 69, 73, 74
                                                            STL: A file format native to the stereolithography CAD
MTT: 2.3.1 Multitouch · 40                                      software · 67
MTT4ALL: Multitouch Table developed by the Author
    of this master thesis · 41, 42
                                                            T

N                                                           Tabletops: Horizontal Interactive Displays · 40, 41


NP: Number of Elements in Each Generation · 34, 36,
                                                            V
    37
NURBS: Non-uniform rational basis spline · 14, 16
                                                            VEs: Virtual Environments · 53
                                                            VR: Virtual Reality · 42, 43, 53, 54, 55
                                                            VRML: Virtual Reality Markup Language · 43




xxxii
xxxiii
I.
                                                          Introduction

1 The Problematic

The adjustment of architectural forms to local and specific solar radiation conditions is a
fundamental study that must be always conducted by architects. When discussing energy
consumption and solar power harness in buildings, important topics of discussion come
into play, like the real relation between a building form and its energy behavior, or finding
the right building shape for a specific location and weather conditions on an all year basis.
Several studies were published so far, to try to answer and demonstrate these and other
important questions. Form follows energy, but how exactly is this happening it‟s somehow
difficult to demonstrate without having automated tools and models. One must try to
manually analyze the energy dependence between form and volume. With this kind of
studies, there is an attempt to simultaneously adapt a building form, in order to increase the
potential areas for solar radiation “reception” and at the same time looks for ways to
reduce the thermal loss (here the admittance method, well known by architects, it‟s useful),
taking in account the need to design for specific locations and specific weather conditions.


This research work aims, to examine the theoretical concepts associated to the problem of
“Form Follows Energy”, pointed out, in studies done by some researchers. Also, the
present study discusses emergent methods based on evolutionary algorithms and
environmental simulation tools and it targets the development of new design methods that
allow the construction of sustainable optimized buildings by using digital technologies,
through the creation of an automated tool and an optimization framework that will allow
the optimization of 3D shapes (buildings), taking in account the geo-location and specific
weather conditions, throughout the run of automated simulations, making autonomous
changes and optimizations utilizing evolutionary algorithms.


                                                                                               1
For the creation of this work, a strong collaboration between the Polytechnic Institute of
Leiria/Superior School of Technology and Management/Centre for Rapid and Sustainable
Product Development (IPL/ESTG/CDRSP) and Fraunhofer Austria was established.


The main research objectives of this work can be listed as follows:


        (i)     To give an overview of state of the art technologies and techniques
                currently employed in the architecture field, regarding simulation and
                analysis, visualization, rendering, virtual interaction, rapid prototyping,
                reverse engineering and automated construction;


        (ii)    To evaluate common shapes used in real world buildings, with focus on
                greenhouses forms as a practical example case study;


        (iii)   To investigate, in order to obtain the necessary parametric equations
                (required to the use of computer graphics in the creation of procedural 3D
                models) for the several identified common building shapes. Also, to
                extract the essential parameters (height, width, length, orientation, roof
                angle…) of those fundamental shapes, in order to achieve a fully
                parametrical defined 3D model, this will allow the use of Euclides
                (JavaScript Procedural Modeler);


        (iv)    To research on the possibility of having a programming integration with
                commonly used simulation packages and tools, to simulate how the
                different shapes of buildings have an influence in energy consumption
                throughout the life of these real buildings, with the final purpose of
                developing an automated tool capable of running automated simulations;




2
(v)     To make use of evolutionary algorithms in order to perform autonomous
                and automatic optimizations of 3D shapes based on automated simulations
                and the well-known method of admittance, always employing tools and
                methods which are widely accepted in the architecture field. The final goal
                is the creation of a global optimization framework for automatic
                generation of optimized 3D building forms, also taking in account the
                specific location weather conditions;


        (vi)    To present and explain the importance of the results achieved with the
                developed global optimization framework, pointing out new directions in
                sustainable architecture design;


Presently, this study is applied to architecture and sustainability. But it must be referred
that this problematic of evolutionary architecture and simulation tools integration can be
extended to other domains/fields, like the industrial or the medical field. They could also
benefit from an autonomous generation of different 3D shapes, as well as a self-governing
optimization of those same 3D forms.




                                                                                           3
1.1 Thesis Structure

The thesis is divided into five chapters, which develops in accordance with the identified
research objectives. This first chapter (Introduction) comprises an introduction that in
addition to listing the key objectives, also briefly describes the context of the research.


The contents of the remaining chapters are summarized as follows:


        State of the Art


         Reviews the latest work developed around procedural modeling, visualization
         techniques, digital fabrication and reverse engineering. It also presents and
         describes a JavaScript Framework, named “Euclides”, utilized for the easy creation
         of 3D procedural and parametric shapes. This was the procedural framework used
         in this thesis for the creation of all the necessary parameterized 3D buildings forms.
         Lastly, a briefly explanation of how evolutionary algorithms work, is also given.
         Moreover, a specific evolutionary algorithm (Differential Evolution - DE) is
         described, as well as the reasons why this particularly algorithm was chosen in this
         thesis, for the development of an automated tool for 3D shapes (buildings forms)
         optimization. Other evolutionary algorithms are pointed out too, as plausible
         alternatives to be implemented within the optimization framework in future work,
         in order to tackle other problematic;


        Simulation Tools in Architecture


         Gives an overview of different interests in simulation, in particular those related to
         the problematic of architecture and energy consumption in buildings. Some
         simulation tools/packages are presented, together with the reasons for selecting a
         particular tool to be used in the work presented here;




4
   A Global Optimization Framework


    Explains the work developed throughout this thesis, on the problematic of how
    buildings form affects the energy consumption on a daily basis throughout its entire
    life. Several methodologies used for choosing parameters to control a specific 3D
    shape as well as other tools used to deduce parametric equations and “mathematical
    code”, are also described with the objective to show how these 3D parametric
    models were generated using Euclides and JavaScript. Also, the general concept of
    the developed optimization framework is explained. A practical overview of the
    work is given, every framework component is presented in more detail and the
    achieved results are presented and clarified. Finally a case study is presented, where
    the problem of automatic optimization is extremely relevant and the results
    obtained are then presented and explained;


   Conclusions and Future Work


    Provides an overall summary of the thesis and points out further progress paths and
    improvement options for the autonomous global optimization framework that was
    developed and presented in this thesis;




                                                                                         5
II.
                                                          State of the
                                                              Art

2 Overview

A review on the state of the art is presented, regarding current work focused on procedural,
parametric and adaptive architecture modeling. A short description of evolutionary
algorithms is given. Also, innovative methods of visualization and presentation of
architecture projects are presented, as well as several techniques for rapid prototyping and
reverse engineering. These methods and techniques are essential to capture 3D geometry,
for achieving more complete results on any architecture project (e.g. production of scale
models for simulation in wind tunnels, virtual simulation, building control…) and essential
for presenting the achieved results to final customers (rendering, interactivity…), also, the
author of this thesis was a research member of the CDRSP Research Centre and earned a
scholarship, on the topic “Build-it-Green”, from the Portuguese Foundation for Science
and Technology (FCT). This “Build-it-Green” topic is closely related to the architecture
subject and some of the work that was developed at CDRSP by the author, was focused on
these same areas and it was conducted throughout the realization of this master thesis.


This state of the art review, aims to present a short explanation about each product,
methodology or recent development. For getting more insightful details, the correspondent
references should be further investigated.




                                                                                            7
2.1 Parametric and Procedural Modeling

Parametric Computer Aided Design (CAD) modeling assumes, nowadays, an important
role in the definition of 3D models. There are several active attempts to collect all the
information about a product or about the different parts that compose a product.
Information platforms like Product Data Management (PDM) or Product Lifecycle
Management (PLM) [1], offer a way to gather the different distributed data that is vital for
an efficient product management. However there is some “intelligent” information that
must be captured with each part and product assembly, such as parametric information
(e.g. width, height, volume, orientation, length, relations between parts, formulas …). Also
semantic methods, using ontologies, try to present solutions for solving problems like the
relationship between different, yet related 3D geometric information [2]. Procedural
modeling can be viewed as the use of different techniques in computer graphics to create
(generate) 3D models and textures from sets of rules. L-Systems, fractals, and generative
modeling are procedural modeling techniques since they apply algorithms for producing
scenes. The set of rules may either be embedded into the algorithm, configurable by
parameters, or a set of rules that is completely separated from the evaluation engine. The
output is then called procedural content, which can be used in computer games, films, be
uploaded to the internet while requiring much less bandwidth, or the user may edit the
content manually [3].


Procedural models often exhibit database amplification, meaning that large scenes can be
generated from a much smaller amount of rules. If the employed algorithm produces the
same output every time, the output needs not to be stored. Often, it is sufficient to start the
algorithm with the same random seed to achieve the same result. Although all modeling
techniques on a computer require algorithms to manage and store data at some point,
procedural modeling focuses on creating a model from a rule set.


Procedural modeling is often applied when it would be too cumbersome to create a 3D
model using generic 3D modelers, or when more specialized tools are required, this is
often the case for plants, architecture or landscapes [4].




8
2.1.1  Plugins for existent 3D Software: Blender, 3D
   Studio Max and Others

There are many plugins available on the internet for use within commonly used 3D
modeling software, like Autodesk 3D Studio Max, Autodesk Maya or the open source
modeling software Blender and many others that allow the automatic generation of terrain,
buildings or even cities in a procedural way.


These plugins permit the creation of 3D models according to specified rules and custom
parameters specified by the user, they can also be customized through the use of scripting
languages like, Python or MEL. Suicidator City Generator (SCG) is a wonderful example
of such plugin for use inside Blender. It is a Python script for Blender or in other words it
is a program written in the Python programming language that runs inside the Blender
environment [5].


It‟s not the purpose of this work to explain in detail how these plugins perform, however
they must be mentioned here as an existent and possible path for the creation of generative
components in today‟s 3D modeling software packages.




                     Fig. II-1 – Suicidator City Generator (SCG) for Blender




                                                                                            9
2.1.2      CityEngine

City Engine (now acquired by ESRI) is one of the most successful and powerful examples
for procedural modeling, it‟s a standalone software that provides a unique conceptual
design and modeling solution for the efficient creation of 3D cities and buildings, for
professional users in entertainment, architecture, urban planning, Geographic Information
Systems (GIS) and general 3D content production [4]. CityEngine was also tested in this
master thesis study.




                  Fig. II-2 – Procedural/parametric examples created in CityEngine


The key highlights of CityEngine include [6]:


        GIS/CAD Data Support and OpenStreet Map Import


         CityEngine supports industry standard formats like, ESRI Shape file or DXF which
         allow the import/export of any geo-spatial/vector data such as parcels, building
         footprints with arbitrary attributes, or line data to create street networks. To copy
         real cities or efficiently create an urban environment for our design, it‟s possible to
         use data from OpenStreet Map. Geospatial data of real cities can also be
         downloaded and directly imported it into CityEngine;




10
   Dynamic City Layouts and Street Networks Patterns


    An intuitive toolset is provided to interactively design, edit and modify urban
    layouts consisting of (curved) streets, blocks and parcels. Street construction or
    block subdivision is controlled via parametric interfaces, giving immediate visual
    feedback; CityEngine offers unique street grow tools to quickly design and
    construct urban layouts. Street patterns such as, grid, organic or circular, are
    available and the topography of the terrain is taken into account;


   Rule-based Modeling Core


    Procedural modeling based on Computer Generated Architecture rules (CGA shape
    grammar) offers unlimited possibilities to control mass, geometry assets,
    proportions, or texturing of buildings or streets on a city-wide scale. We can define
    our own rules using custom textures/models in the node- or text-based rule editor;


   Facade Wizard and Parametric Modeling Interface


    One can quickly create rules out of an image or a textured mass model with this
    simple and easy-to-use visual facade authoring tool. The resulting facade rules are
    size-independent, contain level-of-detail and can be extended with e.g. detailed
    window asset. A convenient interface to interactively control specific street or
    building parameters such as the height or age (defined by the rules) is provided and
    with the live mode, parameter modifications invoke the automatic regeneration of
    the 3D model;


   Map-Controlled City Modeling and Reporting (Building Information Modeling -
    BIM for Cities)


    Any parameter of the buildings and streets can be controlled globally via image
    maps (for example the building heights or the land use-mix); this allows for
    intuitive city modeling and quick changes on a city-wide scale. Furthermore,
    terrains can be imported, aligned, and exported. Customized rule-based reports can
    be generated to analyze the urban design e.g. automatically calculate quantities



                                                                                         11
such as Gross Floor Area (GFA), Floor Area Ratio (FAR), etc. Reports are updated
         automatically and instantaneously and can be made for whole city parts;


        Industry-Standard 3D Formats


         CityEngine supports Collada, Autodesk FBX, 3DS, Wavefront OBJ and e-on
         software's Vue, which allow for flawless 3D data exchange; FBX and Collada
         support asset instancing, multiple UV-sets, grouping and binary encoding;
         furthermore, scenes can also be exported to RenderMan RIB or Mental Ray MI
         format. Textures can be collected during (batch) export;


        Python


         Allows streamlining repetitive or pipeline-specific tasks with the integrated Python
         scripting interface (e.g. write out arbitrary meta-data or instancing information for
         each building, import FBX cameras, etc...). CityEngine is also available for
         Windows (32/64 bits), Mac OSX (64 bits), and Linux (32/64 bits).




                    Fig. II-3 – CityEngine IDE and the Rule Editor Capabilities




12
2.1.3  Bentley – MicroStation Extension:
   GenerativeComponents (GC)

Designers have (since the dawn of times), wanted to innovate. Indeed, innovation is widely
regarded as a trophy that awaits creative professionals who successfully explore endless
design alternatives to ultimately arrive at the most efficient solution - a process that can be
incredibly time consuming as each alternative is thoroughly modeled and assessed. Using
the existing tools, a minor change to a design may require a major update to the model,
thus restricting the number of design alternatives considered by the team due to time
constraints. GenerativeComponents is an associative parametric modeling system used by
architects and engineers to automate design processes and accelerate design iterations. As
an innovation by MicroStation, GenerativeComponents extends proven technologies and
delivers significant advantage to users as they rapidly explore a broad range of design
alternatives. With a hybrid approach, designers who use GenerativeComponents can model
geometry, capture relationships, and generate forms using scripts and/or direct
manipulation for unrivalled creative flexibility [7].




                Fig. II-4 – GenerativeComponents (GC) IDE, Bentley MicroStation


This combination of accelerated iteration, flexible modeling, and automated process,
means that a GenerativeComponents design can be highly efficient, benefiting from a
combination of intuition and logic [7].




                                                                                             13
2.1.4     Rhinoceros and Grasshopper

Rhino (a.k.a. Rhinoceros) is a stand-alone, commercial NURBS-based 3D modeling tool,
developed by Robert McNeel & Associates. The software is commonly used for industrial
design, architecture, marine design, jewelry design, automotive design, CAD / CAM, rapid
prototyping, reverse engineering as well as the multimedia and graphic design industries.




               Fig. II-5 – Rhinoceros is used in multiple fields, including architecture


Rhino is specialized in free-form non-uniform rational B-spline (NURBS) modeling. Plug-
ins    developed   by    McNeel      includes Flamingo (retrace rendering),           Penguin   (non-
photorealistic rendering), Bongo and Brazil (advanced rendering). Over one hundred third-
party plugins are available. There are also rendering plug-ins for Maxwell Render, V-ray
and many other engines. Additional plugins for CAM and CNC milling are available as
well, allowing for tool path generation directly in Rhino. Like many other modeling
applications, Rhino also features a scripting language, based on the Visual Basic language
and an SDK that allows reading and writing Rhino files directly. Rhino 3D gained its
popularity in architectural design in part because of the Grasshopper plug-in for
computational design. Many new avant-garde architects are using parametric modeling
tools, like Grasshopper. Rhino's increasing popularity is based on its diversity, multi-
disciplinary functions, low learning-curve, relatively low cost, and its ability to import and




14
export over 30 file formats, which allows Rhino to act as a “converter” tool between
programs in a design workflow. The combination between Rhino and Grasshopper is just
perfect to create all kinds of parametric studies and developments on any field. The power
offered by Rhino and Grasshopper is just amazing. There are also many other plugins
available (rendering, math, physics, kinematics…).




Fig. II-6 – Voronoi Examples, created using Rhinoceros and Grasshopper, by Atsuo Nakajima (Tokyo,
                                               Japan)


For designers who are exploring new shapes using generative algorithms, Grasshopper is a
graphical algorithm editor tightly integrated with Rhino‟s 3D modeling tools [8]. In Fig.
II-7, by using Grasshopper, a building and its structural supports are generated and
calculated using only two splines created initially in Rhino.




 Fig. II-7 - Parametric Strategies achieved using Rhino and Grasshopper. (Created by the author of
                                          this master thesis)


Unlike RhinoScript, Grasshopper requires no knowledge of programming or scripting, but
still allows designers to build form generators from the simple to the remarkable [8].




                                                                                                     15
2.1.5     Generative Modeling Language (GML)

Traditionally, 3D objects and virtual worlds are defined by lists of geometric primitives:
cubes and spheres in a Constructive Solid Geometry (CSG) tree, NURBS patches a set of
implicit functions, a soup of triangles, or just a cloud of points.


The term “generative modeling” describes a paradigm change in shape description, the
generalization from objects to operations: A shape is described by a sequence of
processing steps, rather than just the end result of applying operations. Shape design
becomes rule design. This approach is very general and it can be applied to any shape
representation that provides a set of generating functions, the “elementary shape
operators”. Its effectiveness has been demonstrated, e.g., in the field of procedural mesh
generation, with Euler operators as complete and closed set of generating functions for
meshes, operating on the half-edge level [9].




Fig. II-8 - Creation of a simple house model using GML, the extrude operator is repeatedly applied to
the ground polygon. To create the roof, the combined operator collapse-mid is applied to the faceCW
                   and faceCCW edges of the edge returned by the extrude operation.


Generative modeling, gains its efficiency through the possibility to create high-level shape
operators from low-level shape operators. Any sequence of processing steps can be
grouped together to create a new “combined operator”. It may use elementary operators, as
well as other combined operators. Concrete values can easily be replaced by parameters
which makes possible the separation of data from operations: the same processing
sequence can be applied to different input data sets. Data can be used to produce different
shapes by applying different combined operators, from, e.g., a library of domain-dependent
modeling operators. This makes possible the creation of very complex objects from only a
few high-level input parameters, such as, a style library [2].




16
Fig. II-9 – Parameterization/Configuration of a Chair with GML


GML is a concrete implementation of the generative approach. Its main feature is that it is
a full functional programming language that can nevertheless be efficiently used as a file
format for low-level shape descriptions. Only 25 Kilobytes GML code of a Gothic window
style library are sufficient to generate connected manifold control meshes for a variety of
windows [10].




                      Fig. II-10 – Gothic Style Building generated with GML


The GML comes with an integrated visualization engine. Thus, it can also be seen as a
viewer with an integrated modeler that overcomes the usual separation of 3D modeling
from interactive visualization. Curved parts are represented as subdivision surfaces that,
within 1-2 seconds, unfold to seven million vertices after four steps of recursive
refinement. The surface is adaptively displayed at interactive rates using optimized
methods for culling and per-face per-frame multi-resolution rendering [11].




                Fig. II-11 – Configuration of Different Wheel Rim Styles using GML




                                                                                         17
2.1.6      Euclides Framework and JavaScript

Enabling an easy access to programming languages that are usually difficult on a direct
approach will dramatically potentiate their use. GML [9] is such a language and can be
described as being similar to Adobe‟s PostScript. A major drawback of all PostScript
dialects is their unintuitive reverse Polish notation, which makes both - reading and writing
- a burdensome task. According to Strobl, M., et al. [12] a language should offer a
structured and intuitive syntax in order to increase efficiency and avoid frustration during
the creation of code. To overcome this issue, Strobl, M., et al. [12] propose a new approach
to translate JavaScript code to GML automatically. Within the last few years generative
modeling techniques have gained attention especially in the context of cultural heritage.
Because a generative model describes a rather ideal object and not a real one, generative
techniques are a basis for object description and classification. This procedural knowledge
differs from other kinds of knowledge, such as declarative knowledge, in a significant way.
It can be applied to a task. This similarity to algorithms is reflected in the way generative
models are designed: they are programmed. In order to make generative modeling
accessible to cultural heritage experts, Schinko, C., et al. [13] created a generative
modeling framework which accounts for their special needs. The result is a generative
modeler called Euclides based on an easy-to-use scripting language (i.e. JavaScript). The
generative model meets the demands on documentation standards and fulfills sustainability
conditions and its integrated meta-modeler approach makes it independent from hardware,
software and platforms.




Fig. II-12 – Example of a 3D application created for this thesis using Euclides and JavaScript. It allows
            the control of several shape parameters on the “Classic Building form example”.



18
2.1.7        Autodesk Revit Architecture 2012

In the latest years, Autodesk made a strong effort in incorporating new technologies (e.g.
multitouch…) and new amazing functionalities (physics, energy analysis, parametric
design) in existing products like Autodesk Revit 2012 or Maya 2012, making a strong
contribution for the development of really innovative products.


Autodesk Revit Architecture can be used to create massing designs; explore design
alternatives based on qualitative and quantitative feedback; and help address various
environmental, constructability, and aesthetic concerns that can arise during project
realization [14].




                    Fig. II-13 – Energy Consumption Study using Autodesk Revit


In the early stages of a design, visualizing a concept in 3D enhances a designer‟s ability to
communicate ideas. Analyzing these ideas yields the ability to predict and optimize the
real-world performance of the built project.


These attributes form a core value of the Building Information Modeling (BIM) process,
which Revit Architecture software is purpose-built to support.




                                                                                           19
In Autodesk Revit Architecture, users have access to a robust collection of easy-to-use
modeling tools that facilitate design conceptualization, visualization, and communication.
This software supports several new modeling operations, including adaptive, component-
driven geometry, robust UV grid manipulation and increased schedule functionality
through reporting parameters. In addition, Revit users on Autodesk Subscription can now
access tools that enable them to better assess the impact of their early design decisions on
energy consumption and carbon emissions without leaving the Revit environment. In order
to clearly illustrate a complete workflow using the conceptual design and analysis tools
and to address the new features introduced with the previous release [14]:


        The Project requirements section outlines the criteria that will drive the building
         design;


        The Parametric Massing Design section describes the steps taken to explore
         massing design alternatives informed by qualitative and quantitative feedback;


        The Site and Environmental Analysis section addresses the impact of building mass
         and orientation on energy consumption and overshadowing;


        The custom “Panelization” section uses the mass design options generated in the
         first section as the basis for informed panel‟s studies.




                    Fig. II-14 – Conceptual Design in Autodesk Revit Architecture




20
2.1.8       Project Vasari and Project Nucleus

Autodesk Project Vasari is an easy-to-use, expressive design tool for creating building
concepts and it‟s build on the same technology as the Autodesk Revit platform.


Project Vasari goes further, with integrated analysis for energy and carbon, providing
design insight where the most important design decisions are made. And, when it‟s time to
move the design to production, simply bring your Vasari design data into the
Autodesk Revit platform for BIM, ensuring clear execution of design intent.


Project Vasari is still under development and is primarily intended to reduce the building
energy loads, not to replace the more detailed analysis tools. It is able to produce
conceptual models using both geometric and parametric modeling functionality. The
designs can be analyzed using the built-in energy modeling and analysis features. The tools
depends on Green Building Studio (Autodesk‟s green building analysis web service) in
many input energy related parameters [15].




                        Fig. II-15 – Sun Studies using project “Vasari”


Project Vasari is focused on conceptual building design using both geometric and
parametric modeling. It supports performance-based design via integrated energy modeling
and analysis features. This new technology preview is now available as a free download
and trial on Autodesk Labs.



                                                                                         21
Project Nucleus integrates the Nucleus simulation engine from Autodesk Maya into
Autodesk Revit Architecture and Project Vasari. It allows designers to experiment with
"form-finding" in the conceptual design phase by simulating forces directly in Revit
Architecture and Project Vasari (the latest technology preview of Project Vasari already
includes the Project Nucleus functionality).




                          Fig. II-16 - Panel Study using Revit and Vasari


Project Nucleus can simulate a wide range of physical phenomena in real time, like wind,
gravity, constraints, and collisions. These forces can help architects generate free-form
shapes, many of which would be impossible to model by hand [16].




        Fig. II-17 - Using Revit, Vasari and Nucleus Physics for a Panel Study, plus Analysis




22
2.1.9          Autodesk Adaptive Components

Adaptive geometry can be sized and positioned in the context where it is used. When you
designate under constrained geometry as adaptive, you specify the geometric elements
allowed to change, while controlling the elements that you want to remain a fixed size or
position [17].


“Adaptivity” is the functionality, within Inventor, that allows the size of a part/feature to be
determined by setting a relationship to another part in an assembly. Basically, “adaptivity”
is a special way to add constraints. These constraints differ from regular constraints in that
they are driven from a separate file. This separate file can be an assembly file or another
part within the assembly file.


A good example of “adaptivity“, is constraining a shaft to a hole in another part. If set up
correctly, when the size of the hole changes the diameter of the shaft updates as well.
“Adaptivity” is normally used during the initial design phase of a model, when changes are
made rapidly and many parts are affected.




                       Fig. II-18 – Adaptive Components in Autodesk Revit




                                                                                              23
Once a design is released, and parts become standard parts, available for use in other
designs, “adaptivity” should be removed to eliminate the possibility of inadvertently
changing a released design. Removing “adaptivity” also improves performance.




                      Fig. II-19 – Adaptive Panel example in Autodesk Revit


As with using any other constraint, forethought should be given to how a design may
change before “adaptivity” is applied. If a part is not likely to change, it is better to apply
normal (non-adaptive) constraints. “Adaptivity” should be used only when absolutely
necessary [17].




          Fig. II-20 – Another Adaptive Panel example, built using Adaptive Components




24
2.2 Evolutionary Architecture and the Use of
  Algorithms in Optimization of Problems

The first references to this field of computation, Evolutionary Solvers or Genetic
Algorithms [18], can be found in the early 60's when Lawrence J. Fogel published the
revolutionary paper "On the Organization of Intellect" [19] which steered the first
endeavors into evolutionary computing. The early 70's saw further ventures with important
work produced by Ingo Rechenberg and John Henry Holland (and others) [20].
Evolutionary Computation didn't gain popularity beyond the programmer world until
Richard Dawkins (one of my favorite authors) came out with the book, "The Blind
Watchmaker" in 1986 [21], which was published with a small program that generated an
apparently endless stream of body-plans called "Bio-morphs" based on human selection.




         Fig. II-21 – Image taken from the book “The Selfish Gene” by Richard Dawkins


After the 80's, the dawn of the personal computer has made it possible for individuals
without government funding to apply evolutionary principles to personal projects and
making it a common jargon. The term "Evolutionary Computing" is very well commonly
known at this time, but is still very much a programmer‟s tool (by programmers and for
programmers) [18, 22].




                                                                                        25
The applications out there that apply evolutionary logic are either aimed at solving specific
problems or they are generic libraries that allow other programmers to develop their own
software [21].




             Fig. II-22 – Several visions related to Evolutionary Architecture and Biomimetic


One of the most important works ever published published about Evolutionary
Architecture was the book of John Fraser, “An Evolutionary Architecture” [23], in the
book introduction one can read: “…in this book the author investigates the fundamental
form-generating processes in architecture, considering architecture as a form of artificial
life, and proposing a genetic representation in a form of DNA-like code-script, which can
then be subject to developmental and evolutionary processes in response to the user and the
environment. The aim of an evolutionary architecture is to achieve in the built
environment, the symbiotic behavior and metabolic balance found in the natural
environment. To do so, it operates like an organism, in a direct analogy with the underlying
design process of nature”.




     Fig. II-23 – Evolutionary examples taken from the book “An Evolutionary Architecture” by John
                                                Fraser [23]



26
Also, Gordon Pask wrote on his foreword on this same book: “The book also proposes a
fundamental change in practice… „The role of the architect here, I think, is not so much to
design a building or city as to catalyze them: to act that they may evolve‟. Promising
sustainable design methods are unquestionably emerging through the use of evolutionary
computation and environmental simulation tools, as this is indeed an essential need in
today‟s architecture world”.




Fig. II-24 – Dynamic Geometry Computation, “Shanghai Tower - Geometry Generate and Rendering”
                                        (Michael Peng)


Eugene Tsui on his work and book “Evolutionary Architecture: Nature as a Basis for
Design” [24] and also Javier Senosiain, Michael Paulin, William McDonough, Renzo
Piano and many others architects incorporate in their projects ecological and sustainable
principles, but also integrate an understanding that constructions require “an holistic
approach studying the form, materials and efficiency that Nature have becoming the
infallible mentor in the creation of an comfortable and symbiotic world” [25].




                    Fig. II-25 – Ecological House of the Future (Eugene Tsui)




                                                                                           27
2.2.1     Differential Evolution (DE)

Differential Evolution (DE) [26] has been very successful in solving the global continuous
optimization problem [27]. It mainly uses the distance and direction information from the
current population to guide its further search. The global optimization problem arises in
almost every field of science, engineering or business, and an enormous amount of effort
have been devoted to solving this problem. The major challenge of the global continuous
optimization is that the problems to be optimized may have many local optima (Sun, J., et
al. [27]).




 Fig. II-26 – The Global Optimization Problem (example in Matlab). Search of the highest mountain
                     peak among a neighborhood of other high mountains peaks.


Evolutionary Algorithms (EA‟s) are similar to the evolution process of a biological
population which can adapt to the changing environments in order to find the optimum of
the optimization problem by evolving a population of candidate solutions. Differential
Evolution (DE) is one of the most successful EAs for the global continuous optimization
problem. Several examples of problem solving using DE were already presented in the
past, particularly those ones presented by the creators of the DE algorithm (Price, K. S., et
al., [28]).




28
Evan Greenberg [29] discusses in is master thesis a natural behavior called “Branching”
that occurs in natural systems for functional reasons. The branching logic for each specific
system is quite different due to environmental and mathematical factors. In the
computation of branching systems, these mathematical factors can be incorporated easily
into the coding of each system. Nevertheless, the environmental components deserve
further consideration in the simulation of these natural systems. Through the engine of
genetic algorithms based on evolutionary developmental theory, the specific logics
observed and analyzed in branching patterns of river systems or trees, can be simulated and
optimized in a digital environment.




  Fig. II-27 - Observation, Analysis and Computation of Branching Patterns in Natural Systems (by
                                        Evan Greenberg [29])


There are some biological terminologies which are used in evolutionary algorithm
implementations, such as:


      Individual: an autonomous piece characterized by a chromosome. In this case, one
       possible solution to the design problem;


      Population: a group of individuals;


      Population Size: the number of individuals in a population Gene (a functional block
       of DNA);




                                                                                                    29
   Allele: A possible value of a gene;


        Chromosome: Strings of DNA. In this case, a list of parameters;


        Locus: The place of a gene in a chromosome.


In evolutionary algorithms there are also three different types of operators: Selection,
Crossover, and Mutation. After initializing the parameters, these three operators are
iterated until the results satisfy the terminal criteria defined.


Each step of the algorithms is explained as follows (Kawakita, G. [30]):


        Initialization - In this step, some parameters including the population size, number
         of generations and so on are entered. After that, the initial input randomly generates
         genotype individuals of the first generation. Particularly, population size is
         significant in terms of the operations, the lengthier the chromosome length is, the
         bigger the population size is. Additionally, a bigger population size requires longer
         calculation time until convergence. However, small population sizes may result in
         premature and undesirable convergence;


        Evaluation - Fitness scores are calculated for further selection of fitter
         chromosomes. One of the most important aspects in this step is the Fitness Function
         which calculates the fitness measurement of each individual. This operation is
         deeply related to the efficiency of the whole Genetic Algorithm (GA) flow;
         therefore, it needs to be determined carefully;


        Selection - The fitter chromosomes in the population are basically selected for
         reproduction. As in biological evolution, the fitter chromosomes are more likely to
         be selected and reproduced in each generation. Meanwhile, lower fitness
         chromosomes are also possibly selected, but with a lower probability. This
         probabilistic selection depends on the selection method. There are several types of
         selection such as elite selection, roulette selection, tournament selection, etc. Each
         selection type has advantages and disadvantages. For instance, in elite selection, the




30
fitter chromosomes are certainly selected in order; however, premature
    convergence is highly possible;


   Crossover - Crossover roughly mimics the genetic operation of biological
    recombination between two chromosomes. The fitter chromosomes are chosen by
    the selection operator. However, it is not effective enough to evolve the population.
    The crossover operator encourages more variation by exchanging genes between
    two chromosomes;


   Mutation - The mutation operator randomly flips or changes genes in a
    chromosome between alleles, generally with a very low probability. Chromosomes
    generated by the crossover operator are basically copies of the parent
    chromosomes. Therefore, premature convergence possibly occurs. Chromosomes
    that have been mutated help to avoid premature convergence. Generally speaking,
    the mutation rate should be 1/L, where L is the length of chromosome. Moreover, if
    the mutation rate is too big, the algorithm becomes similar to a random search;


   Terminal Criterion - In this step, the conditions required to terminate the algorithm
    is evaluated. If the process is regarded as being completed, the fittest individual in
    the generation is outputted as one of the possible optimum solutions.


The general conditions of convergence in evolutionary algorithms are as follows:


       If the fittest score in the population satisfies the certain target – star gene;


       If the average fitness score in the population satisfies the certain target –
        population improvement;


       If the increase or decrease of fitness scores in the population becomes below a
        certain value – convergence;


       If the number of generations becomes over the defined value – finite iteration.




                                                                                           31
Fig. II-28 – Simple EA’s steps




As it happens with every algorithm, there are several different variations of the differential
algorithm, in order to classify the different variants, the notation: DE/x/y/z was introduced,
where:


        x specifies the vector to be mutated which can be “rand” (a randomly chosen
         population vector) or “best” (the vector of lowest cost from the current population);


        y is the number of difference vectors used;


        z denotes the crossover scheme. Example: “bin” (Crossover due to independent
         binomial experiments).


Using this notation, the basic DE-strategy that is generally described can be written as:
DE/rand/1/bin, but there are other variants, e.g. DE/best/2/bin.



32
The DE algorithm work‟s (in general) as follows (Price, K. S., et al.[28]):


    1. The DE algorithm maintains a population of N points in every generation, where
        each point is a potential solution and N is a control parameter;


    Then the algorithm evolves and improves the population iteratively:


    2. In each generation, a new population is generated based on the current population;


    3. To generate descendants for the new population, the algorithm extracts distance and
        direction information from the current population members and adds random
        deviation for achieving diversity;


    4. If an offspring has a lower objective function value than a predetermined
        population member, it will replace this population member;


    5. This evolution process continues until a stopping criterion is met (e.g., the current
        best objective function value is smaller than a given value or the number of
        generations is equal to a given maximum value).




 Fig. II-29 – General Evolutionary Algorithm: i: initialization, f(X): evaluation, ?: stopping criterion,
Se: selection, Cr: cross-over, Mu: mutation, Re: replacement, X*: optimum. Author: Johann "nojhan"
                                                  Dréo




                                                                                                       33
The optimization method known as Differential Evolution (DE) has several parameters that
determine its behavior and efficacy in optimizing a given problem. The selection of good
parameters for DE it is an important question that is discussed by Pedersen, M. [31], this
paper gives a list of good possible choices of parameters values for various optimization
scenarios with the intention to give an easy help when choosing the best values for
achieving the best results, these interrelated parameters are: Crossover (CR), usually a
good initial value for CR would be 0.9 or 1.0 to check if a quick solution is possible,
Number of evaluations (Fitness Evaluations), Number of elements in each generation (NP),
a good value for NP is between 5*D (D = Dimension of the problem) and 10*D, but NP
must be at least 4 to ensure that DE will have enough mutually different vectors to which
to work on, Differential Weight (F), a good initial value for F is usually 0.5, Number of
variables of the problem (Problem Dimensions) and Size of the domain for each variable of
the problem.




     Fig. II-30 - DE optimization performance (Pedersen, M. [31]) on several different problems using
        DE/rand/1/bin algorithm. Plots show the mean fitness achieved. Over 50 optimization runs




34
In simple terms, optimization is the attempt to maximize a system‟s desirable properties
while simultaneously minimizing its undesirable characteristics. What these properties are
and how effectively they can be improved depends on the problem at hand (Price, K. S., et
al. [32]).


The optimization method known as Differential Evolution (or DE) was originally
introduced by Storn and Price [32] and offers a way of optimizing a problem without using
its gradient. This is particularly useful if the gradient is difficult or even impossible to
derive (Pedersen, M. [31]). DE maintains a population of agents which are iteratively
combined and updated using simple formula to form new agents. The general purpose
optimization method known as DE has a number of parameters that determine its behavior
and efficacy in optimizing a given problem, these parameters must be chosen accordingly
[31, 32]. Small changes to the DE implementation can cause dramatic changes in the
behavioral parameters that cause good optimization performance. The parameters given by
Pedersen [31] have been tuned for the DE/rand/1/bin algorithm. If other DE
implementation is chosen, different parameters values must be selected.


According to Storn and Price [26], the DE algorithm (DE/rand/1/bin) was demonstrated to
converge faster and with more certainty than many other acclaimed global optimization
methods. DE also requires fewer control variables, it‟s robust, easier to use and fits itself
on parallel computation scenarios. Storn and Price [26] compared and tested DE against
other optimization methods in solving minimization problems. The DE algorithm uses
simultaneous search vectors in order to help escape local minima (not using only a usual
greedy criterion, where a new parameter vector is accepted if an only if it reduces the value
of the cost function).


Also DE was compared to Simulated Annealing [33] (annealing relaxes the greedy
criterion by occasionally permitting an uphill move, allowing a parameter vector to climb
out of a local minimum, however in a long run this method also leads to a greedy
criterion).




                                                                                           35
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms
Optimal Forms

Contenu connexe

En vedette

How To Get More From SlideShare - Super-Simple Tips For Content Marketing
How To Get More From SlideShare - Super-Simple Tips For Content MarketingHow To Get More From SlideShare - Super-Simple Tips For Content Marketing
How To Get More From SlideShare - Super-Simple Tips For Content MarketingContent Marketing Institute
 
A Guide to SlideShare Analytics - Excerpts from Hubspot's Step by Step Guide ...
A Guide to SlideShare Analytics - Excerpts from Hubspot's Step by Step Guide ...A Guide to SlideShare Analytics - Excerpts from Hubspot's Step by Step Guide ...
A Guide to SlideShare Analytics - Excerpts from Hubspot's Step by Step Guide ...SlideShare
 
2015 Upload Campaigns Calendar - SlideShare
2015 Upload Campaigns Calendar - SlideShare2015 Upload Campaigns Calendar - SlideShare
2015 Upload Campaigns Calendar - SlideShareSlideShare
 
What to Upload to SlideShare
What to Upload to SlideShareWhat to Upload to SlideShare
What to Upload to SlideShareSlideShare
 
How to Make Awesome SlideShares: Tips & Tricks
How to Make Awesome SlideShares: Tips & TricksHow to Make Awesome SlideShares: Tips & Tricks
How to Make Awesome SlideShares: Tips & TricksSlideShare
 
Getting Started With SlideShare
Getting Started With SlideShareGetting Started With SlideShare
Getting Started With SlideShareSlideShare
 

En vedette (6)

How To Get More From SlideShare - Super-Simple Tips For Content Marketing
How To Get More From SlideShare - Super-Simple Tips For Content MarketingHow To Get More From SlideShare - Super-Simple Tips For Content Marketing
How To Get More From SlideShare - Super-Simple Tips For Content Marketing
 
A Guide to SlideShare Analytics - Excerpts from Hubspot's Step by Step Guide ...
A Guide to SlideShare Analytics - Excerpts from Hubspot's Step by Step Guide ...A Guide to SlideShare Analytics - Excerpts from Hubspot's Step by Step Guide ...
A Guide to SlideShare Analytics - Excerpts from Hubspot's Step by Step Guide ...
 
2015 Upload Campaigns Calendar - SlideShare
2015 Upload Campaigns Calendar - SlideShare2015 Upload Campaigns Calendar - SlideShare
2015 Upload Campaigns Calendar - SlideShare
 
What to Upload to SlideShare
What to Upload to SlideShareWhat to Upload to SlideShare
What to Upload to SlideShare
 
How to Make Awesome SlideShares: Tips & Tricks
How to Make Awesome SlideShares: Tips & TricksHow to Make Awesome SlideShares: Tips & Tricks
How to Make Awesome SlideShares: Tips & Tricks
 
Getting Started With SlideShare
Getting Started With SlideShareGetting Started With SlideShare
Getting Started With SlideShare
 

Similaire à Optimal Forms

A BIM-integrated approach to construction quality management
A BIM-integrated approach to construction quality managementA BIM-integrated approach to construction quality management
A BIM-integrated approach to construction quality managementEsper Achkar
 
Costofbadpm schiltz v11
Costofbadpm schiltz v11Costofbadpm schiltz v11
Costofbadpm schiltz v11Referendo Org
 
6.1 method for system design for sustainability
6.1 method for system design for sustainability6.1 method for system design for sustainability
6.1 method for system design for sustainabilityLeNS_slide
 
7.1 design exercise presentation 10 11 (47)
7.1 design exercise presentation 10 11 (47)7.1 design exercise presentation 10 11 (47)
7.1 design exercise presentation 10 11 (47)LeNS_slide
 
7.1 design exercise presentation 10 11 (47)
7.1 design exercise presentation 10 11 (47)7.1 design exercise presentation 10 11 (47)
7.1 design exercise presentation 10 11 (47)LeNS_slide
 
bringing design to life with lean ux & lean engineering - Lean Day West 2013
bringing design to life with  lean ux & lean engineering - Lean Day West 2013bringing design to life with  lean ux & lean engineering - Lean Day West 2013
bringing design to life with lean ux & lean engineering - Lean Day West 2013Bill Scott
 
Report[Batch-08].pdf
Report[Batch-08].pdfReport[Batch-08].pdf
Report[Batch-08].pdf052Sugashk
 
0.0 Sds Course Introduction Vezzoli 07 08 (28.10.08)
0.0 Sds Course Introduction Vezzoli 07 08 (28.10.08)0.0 Sds Course Introduction Vezzoli 07 08 (28.10.08)
0.0 Sds Course Introduction Vezzoli 07 08 (28.10.08)vezzoli
 
6.1 design exercise presentation 09 10 (mo 24 am) (44)
6.1 design exercise presentation 09 10 (mo 24 am) (44)6.1 design exercise presentation 09 10 (mo 24 am) (44)
6.1 design exercise presentation 09 10 (mo 24 am) (44)vezzoliDSS
 
Contemporary Software Engineering Practices Together With Enterprise
Contemporary Software Engineering Practices Together With EnterpriseContemporary Software Engineering Practices Together With Enterprise
Contemporary Software Engineering Practices Together With EnterpriseKenan Sevindik
 
From Developer and Beyond - The IT Architect Career
From Developer and Beyond - The IT Architect CareerFrom Developer and Beyond - The IT Architect Career
From Developer and Beyond - The IT Architect CareerMarcelo Sousa Ancelmo
 
6.1 method for system design for sustainability vezzoli 14-15 (71)
6.1 method for system design for sustainability vezzoli 14-15 (71)6.1 method for system design for sustainability vezzoli 14-15 (71)
6.1 method for system design for sustainability vezzoli 14-15 (71)LeNS_slide
 
6.1 method for system design for sustainability vezzoli 14-15 (71)
6.1 method for system design for sustainability vezzoli 14-15 (71)6.1 method for system design for sustainability vezzoli 14-15 (71)
6.1 method for system design for sustainability vezzoli 14-15 (71)Emanuela Emy
 
Organisering av digitale prosjekt: Hva har IT-bransjen lært om store prosjekter?
Organisering av digitale prosjekt: Hva har IT-bransjen lært om store prosjekter?Organisering av digitale prosjekt: Hva har IT-bransjen lært om store prosjekter?
Organisering av digitale prosjekt: Hva har IT-bransjen lært om store prosjekter?Torgeir Dingsøyr
 
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...maheshshinde762539
 
Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi
Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-EdisiProduct-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi
Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-EdisiSreesh P Somarajan
 
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...IrumSaba11
 

Similaire à Optimal Forms (20)

Sanni Siltanen: Developing augmented reality solutions through user involveme...
Sanni Siltanen: Developing augmented reality solutions through user involveme...Sanni Siltanen: Developing augmented reality solutions through user involveme...
Sanni Siltanen: Developing augmented reality solutions through user involveme...
 
Annual Report
Annual ReportAnnual Report
Annual Report
 
A BIM-integrated approach to construction quality management
A BIM-integrated approach to construction quality managementA BIM-integrated approach to construction quality management
A BIM-integrated approach to construction quality management
 
Costofbadpm schiltz v11
Costofbadpm schiltz v11Costofbadpm schiltz v11
Costofbadpm schiltz v11
 
6.1 method for system design for sustainability
6.1 method for system design for sustainability6.1 method for system design for sustainability
6.1 method for system design for sustainability
 
7.1 design exercise presentation 10 11 (47)
7.1 design exercise presentation 10 11 (47)7.1 design exercise presentation 10 11 (47)
7.1 design exercise presentation 10 11 (47)
 
7.1 design exercise presentation 10 11 (47)
7.1 design exercise presentation 10 11 (47)7.1 design exercise presentation 10 11 (47)
7.1 design exercise presentation 10 11 (47)
 
bringing design to life with lean ux & lean engineering - Lean Day West 2013
bringing design to life with  lean ux & lean engineering - Lean Day West 2013bringing design to life with  lean ux & lean engineering - Lean Day West 2013
bringing design to life with lean ux & lean engineering - Lean Day West 2013
 
TR1643
TR1643TR1643
TR1643
 
Report[Batch-08].pdf
Report[Batch-08].pdfReport[Batch-08].pdf
Report[Batch-08].pdf
 
0.0 Sds Course Introduction Vezzoli 07 08 (28.10.08)
0.0 Sds Course Introduction Vezzoli 07 08 (28.10.08)0.0 Sds Course Introduction Vezzoli 07 08 (28.10.08)
0.0 Sds Course Introduction Vezzoli 07 08 (28.10.08)
 
6.1 design exercise presentation 09 10 (mo 24 am) (44)
6.1 design exercise presentation 09 10 (mo 24 am) (44)6.1 design exercise presentation 09 10 (mo 24 am) (44)
6.1 design exercise presentation 09 10 (mo 24 am) (44)
 
Contemporary Software Engineering Practices Together With Enterprise
Contemporary Software Engineering Practices Together With EnterpriseContemporary Software Engineering Practices Together With Enterprise
Contemporary Software Engineering Practices Together With Enterprise
 
From Developer and Beyond - The IT Architect Career
From Developer and Beyond - The IT Architect CareerFrom Developer and Beyond - The IT Architect Career
From Developer and Beyond - The IT Architect Career
 
6.1 method for system design for sustainability vezzoli 14-15 (71)
6.1 method for system design for sustainability vezzoli 14-15 (71)6.1 method for system design for sustainability vezzoli 14-15 (71)
6.1 method for system design for sustainability vezzoli 14-15 (71)
 
6.1 method for system design for sustainability vezzoli 14-15 (71)
6.1 method for system design for sustainability vezzoli 14-15 (71)6.1 method for system design for sustainability vezzoli 14-15 (71)
6.1 method for system design for sustainability vezzoli 14-15 (71)
 
Organisering av digitale prosjekt: Hva har IT-bransjen lært om store prosjekter?
Organisering av digitale prosjekt: Hva har IT-bransjen lært om store prosjekter?Organisering av digitale prosjekt: Hva har IT-bransjen lært om store prosjekter?
Organisering av digitale prosjekt: Hva har IT-bransjen lært om store prosjekter?
 
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...
 
Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi
Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-EdisiProduct-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi
Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi
 
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...
202-Product-Design-and-Development-Karl-T.-Ulrich-Steven-D.-Eppinger-Edisi-6-...
 

Dernier

🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘RTylerCroy
 
Real Time Object Detection Using Open CV
Real Time Object Detection Using Open CVReal Time Object Detection Using Open CV
Real Time Object Detection Using Open CVKhem
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonAnna Loughnan Colquhoun
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationMichael W. Hawkins
 
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking MenDelhi Call girls
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsMaria Levchenko
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slidevu2urc
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Drew Madelung
 
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEarley Information Science
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024The Digital Insurer
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountPuma Security, LLC
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slidespraypatel2
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreternaman860154
 
Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024The Digital Insurer
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Enterprise Knowledge
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024Results
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityPrincipled Technologies
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...apidays
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024The Digital Insurer
 

Dernier (20)

🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘
 
Real Time Object Detection Using Open CV
Real Time Object Detection Using Open CVReal Time Object Detection Using Open CV
Real Time Object Detection Using Open CV
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day Presentation
 
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed texts
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slide
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path Mount
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slides
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreter
 
Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivity
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024
 

Optimal Forms

  • 1. Master Thesis (Dissertação de Mestrado) Master in Design and Product Development Engineering (Mestrado em Engenharia da Concepção e Desenvolvimento de Produto) Optimal Forms Generative Modeling Techniques in Optimization Nelson de Jesus Silvério da Silva Leiria, July 2011
  • 2.
  • 3. Master Thesis (Dissertação de Mestrado) Master in Design and Product Development Engineering (Mestrado em Engenharia da Concepção e Desenvolvimento de Produto) Optimal Forms Generative Modeling Techniques in Optimization Nelson de Jesus Silvério da Silva Scientific Adviser: Dr. Nuno Alves (Escola Superior de Tecnologia e Gestão do Instituto Politécnico de Leiria, Portugal) Scientific Co-Adviser: Dr. Eva Eggeling (Fraunhofer, Austria) Leiria, July 2011
  • 4.
  • 5. Report submitted to the Polytechnic Institute of Leiria in partial fulfillment of the requirements for the degree of Master in Design and Product Development Engineering (Mestrado em Engenharia da Concepção e Desenvolvimento de Produto). ISBN: © Polytechnic Institute of Leiria i
  • 6. ii
  • 8. iv
  • 10. vi
  • 11. Acknowledgments To my Scientific Advisor, Prof. Dr. Nuno Alves (Vice-director of CDRSP Research Centre), for the constant attention, motivation and ideas given throughout the course of the thesis and for helping me in achieving success. Thank you also for your friendship and whole-hearted smile. To my Scientific Co-Adviser, Dr. Eva Eggeling (Business Unit Manager of the Visual Computing Fraunhofer Austria Research GmbH), thank you so much for the support and incentive given to me personally, with your warm friendship and also for the support given professionally that made the realization of this thesis possible. To M.Sc. Torsten Ullrich (Researcher at Fraunhofer, Austria), for priceless knowledge, contribution, helps and of course, he‟s sincere friendship, this was crucial for the development of this innovative work. To Prof. Dr. Paulo Bártolo (Director of CDRSP Research Centre) and Prof. Dr. Helena Bártolo (CDRSP Research Centre) for all the support throughout the course of the thesis and for receiving me at the CDRSP Research Centre. To Prof. Dr. Dieter Fellner (Director of the Fraunhofer Institute for Computer Graphics Research - IGD) for giving me the opportunity to develop this work with the CGV Group and Fraunhofer Austria and for allowing me to have access to all the materials and valuable information produced within the CGV/Fraunhofer group. To M.Sc. Volker Settgast (Researcher at Fraunhofer Austria) for all the great tips about Autodesk Maya and Rendering in general and he‟s invaluable friendship. To Dr. Christina Lemke (Architect with projects in Germany and Spain, urban planner, construction biologist & ecologist) for her support, that kindly allowed me to have access to her published PhD thesis work. To my wife, for her patience, for being always there, supporting and encouraging and for understanding all the time I wasn‟t around. To ALL of you, that made this thesis possible, with your comprehension, motivation, support and encouragement. This master thesis was only possible to achieve, due to the good relationships and close partnership that was settled between CDRSP and Fraunhofer Austria. Thank you all so much… vii
  • 12. viii
  • 13. Keywords Procedural, Optimization, Evolutionary, Algorithm, Simulation, Building Abstract The generative modeling paradigm is a shift from static models to flexible models. A generative model describes a modeling process using functions, methods and operators. The result is an algorithmic description of the construction process. Each evaluation of such an algorithm creates a model instance, which depends on its input parameters (width, height, radius, orientation, etc.). These values are normally chosen according to aesthetic aspects and style. In this study, the model‟s parameters are automatically generated according to an objective function. A generative model can be optimized according to its parameters, in this way, the best solution for a constrained problem is determined. The field of application is energy and architecture. Besides the establishment of an overall framework design, this work consists on the identification of different building shapes and their main parameters, the creation of an algorithmic description for these main shapes and the formulation of the objective function, respecting a building‟s energy consumption (solar energy, heating and insulation). Also, this work aims the conception of an optimization pipeline, combining an energy calculation tool with a geometric scripting engine. In this study, one can read about state of the art developments related to architecture and procedural modeling. The major contribution of this development is to present methods that lead to an automated and optimized 3D shape generation for the projected building (based on the desired conditions and according to specific constrains), this will help in the construction of real buildings that account for less energy consumption and for a more sustainable world. ix
  • 14. x
  • 16. xii
  • 17. Acknowledgments ......................................................................................................................................... vii Keywords ........................................................................................................................................................ ix Abstract.......................................................................................................................................................... ix Table of Contents....................................................................................................................................... xi Figures List ...............................................................................................................................................xvii Tables List ................................................................................................................................................ xxv Abbreviations and Acronyms .................................................................................................................. xxix I. Introduction ......................................................................................................................................... 1 1 The Problematic .................................................................................................................................. 1 1.1 Thesis Structure ............................................................................................................................... 4 II. State of the Art ................................................................................................................................. 7 2 Overview ............................................................................................................................................. 7 2.1 Parametric and Procedural Modeling .............................................................................................. 8 2.1.1 Plugins for existent 3D Software: Blender, 3D Studio Max and Others ....................................... 9 2.1.2 CityEngine .................................................................................................................................. 10 2.1.3 Bentley – MicroStation Extension: GenerativeComponents (GC).............................................. 13 2.1.4 Rhinoceros and Grasshopper ..................................................................................................... 14 2.1.5 Generative Modeling Language (GML) ...................................................................................... 16 2.1.6 Euclides Framework and JavaScript ........................................................................................... 18 2.1.7 Autodesk Revit Architecture 2012 ............................................................................................. 19 2.1.8 Project Vasari and Project Nucleus ............................................................................................ 21 2.1.9 Autodesk Adaptive Components ............................................................................................... 23 2.2 Evolutionary Architecture and the Use of Algorithms in Optimization of Problems ..................... 25 2.2.1 Differential Evolution (DE) ......................................................................................................... 28 2.2.2 Pros and Cons of using Evolutionary Algorithms (EAs) .............................................................. 38 2.2.3 Other Evolutionary Based Algorithms - DE/EDA and Hybrid-DE ................................................ 39 2.3 Advanced Rendering, Visualization and Interaction Techniques in Architecture ......................... 40 2.3.1 Multitouch (MTT) ....................................................................................................................... 40 2.3.2 Virtual and Augmented Reality .................................................................................................. 43 2.3.3 Computer Generated Holography ............................................................................................. 47 2.3.4 Advanced Rendering .................................................................................................................. 50 xiii
  • 18. 2.3.5 Virtual World Interactivity ......................................................................................................... 53 2.4 “A World Full of Sensors” .............................................................................................................. 57 2.4.1 Sensors Feed Information into Virtual Worlds .......................................................................... 57 2.4.2 Remote Monitoring of Persons Inside Buildings ........................................................................ 60 2.4.3 Kinetic, Responsive Performative and Adaptive Architecture ................................................... 61 2.5 Reverse Engineering and Rapid Prototyping ................................................................................. 62 2.5.1 Reverse Engineering................................................................................................................... 63 2.5.2 Rapid Prototyping ...................................................................................................................... 68 2.5.3 Rapid Prototyping Techniques ................................................................................................... 69 2.5.4 Personal Fabrication and Future Manufacturing ....................................................................... 76 2.6 Building Information Modeling (BIM) and Automated Construction of Buildings ........................ 78 2.6.1 Building Information Modeling (BIM) ........................................................................................ 78 2.6.2 Automated Construction of Buildings ........................................................................................ 79 III. Simulation Tools in Architecture ..................................................................................................... 81 3 Outline............................................................................................................................................... 81 3.1 EnergyPlus and DesignBuilder ....................................................................................................... 82 3.2 Autodesk Ecotect Analysis ............................................................................................................. 84 3.2.1 Short Comparison between Autodesk Ecotect and EnergyPlus ................................................. 87 3.3 Ansys: AirFlow ............................................................................................................................... 88 3.4 Sustainability Tools in Architecture – Comparison Studies/Audits ............................................... 90 IV. A Global Optimization Framework ................................................................................................. 91 4 Problematic of “Form Follows Energy” and the Pursuit of Solutions................................................ 91 4.1 Answer to the Problematic: Optimal Forms - A Global Optimization Framework ........................ 97 4.2 Identification of Essential Forms Used in the Real World ............................................................. 98 4.2.1 Procedural Shape Generation .................................................................................................... 99 4.2.2 Code Writing Using Euclides and JavaScript ............................................................................ 101 4.3 Simulation Tools Integration ....................................................................................................... 102 4.3.1 Simulation in Ecotect and the Admittance Method................................................................. 104 4.3.2 Initial Manual Workflow Tests ................................................................................................. 105 4.4 Differential Evolution................................................................................................................... 107 4.5 The Developed Global Optimization Framework ........................................................................ 110 4.6 Case Studies and Presentation of Results.................................................................................... 112 4.6.1 Case Study 1 – Classic Shape Building Optimization ................................................................ 113 4.6.2 Case Study 1 - Presentation of Results..................................................................................... 116 4.6.3 Case Study 2 – Cube Shape Building Optimization .................................................................. 126 xiv
  • 19. V. Conclusions and Future Work ....................................................................................................... 131 5 Summary ......................................................................................................................................... 131 5.1 Final Conclusions ......................................................................................................................... 132 5.2 Future Work................................................................................................................................. 133 REFERENCES ............................................................................................................................................ 135 xv
  • 20. xvi
  • 21. Figures List xvii
  • 22. xviii
  • 23. FIG. II-1 – SUICIDATOR CITY GENERATOR (SCG) FOR BLENDER 9 FIG. II-2 – PROCEDURAL/PARAMETRIC EXAMPLES CREATED IN CITYENGINE 10 FIG. II-3 – CITYENGINE IDE AND THE RULE EDITOR CAPABILITIES 12 FIG. II-4 – GENERATIVECOMPONENTS (GC) IDE, BENTLEY MICROSTATION 13 FIG. II-5 – RHINOCEROS IS USED IN MULTIPLE FIELDS, INCLUDING ARCHITECTURE 14 FIG. II-6 – VORONOI EXAMPLES, CREATED USING RHINOCEROS AND GRASSHOPPER, BY ATSUO NAKAJIMA (TOKYO, JAPAN) 15 FIG. II-7 - PARAMETRIC STRATEGIES ACHIEVED USING RHINO AND GRASSHOPPER. (CREATED BY THE AUTHOR OF THIS MASTER THESIS) 15 FIG. II-8 - CREATION OF A SIMPLE HOUSE MODEL USING GML, THE EXTRUDE OPERATOR IS REPEATEDLY APPLIED TO THE GROUND POLYGON. TO CREATE THE ROOF, THE COMBINED OPERATOR COLLAPSE-MID IS APPLIED TO THE FACECW AND FACECCW EDGES OF THE EDGE RETURNED BY THE EXTRUDE OPERATION. 16 FIG. II-9 – PARAMETERIZATION/CONFIGURATION OF A CHAIR WITH GML 17 FIG. II-10 – GOTHIC STYLE BUILDING GENERATED WITH GML 17 FIG. II-11 – CONFIGURATION OF DIFFERENT WHEEL RIM STYLES USING GML 17 FIG. II-12 – EXAMPLE OF A 3D APPLICATION CREATED FOR THIS THESIS USING EUCLIDES AND JAVASCRIPT. IT ALLOWS THE CONTROL OF SEVERAL SHAPE PARAMETERS ON THE “CLASSIC BUILDING FORM EXAMPLE”. 18 FIG. II-13 – ENERGY CONSUMPTION STUDY USING AUTODESK REVIT 19 FIG. II-14 – CONCEPTUAL DESIGN IN AUTODESK REVIT ARCHITECTURE 20 FIG. II-15 – SUN STUDIES USING PROJECT “VASARI” 21 FIG. II-16 - PANEL STUDY USING REVIT AND VASARI 22 FIG. II-17 - USING REVIT, VASARI AND NUCLEUS PHYSICS FOR A PANEL STUDY, PLUS ANALYSIS 22 FIG. II-18 – ADAPTIVE COMPONENTS IN AUTODESK REVIT 23 FIG. II-19 – ADAPTIVE PANEL EXAMPLE IN AUTODESK REVIT 24 FIG. II-20 – ANOTHER ADAPTIVE PANEL EXAMPLE, BUILT USING ADAPTIVE COMPONENTS 24 FIG. II-21 – IMAGE TAKEN FROM THE BOOK “THE SELFISH GENE” BY RICHARD DAWKINS 25 FIG. II-22 – SEVERAL VISIONS RELATED TO EVOLUTIONARY ARCHITECTURE AND BIOMIMETIC 26 FIG. II-23 – EVOLUTIONARY EXAMPLES TAKEN FROM THE BOOK “AN EVOLUTIONARY ARCHITECTURE” BY JOHN FRASER [23] 26 FIG. II-24 – DYNAMIC GEOMETRY COMPUTATION, “SHANGHAI TOWER - GEOMETRY GENERATE AND RENDERING” (MICHAEL PENG) 27 FIG. II-25 – ECOLOGICAL HOUSE OF THE FUTURE (EUGENE TSUI) 27 FIG. II-26 – THE GLOBAL OPTIMIZATION PROBLEM (EXAMPLE IN MATLAB). SEARCH OF THE HIGHEST MOUNTAIN PEAK AMONG A NEIGHBORHOOD OF OTHER HIGH MOUNTAINS PEAKS. 28 FIG. II-27 - OBSERVATION, ANALYSIS AND COMPUTATION OF BRANCHING PATTERNS IN NATURAL SYSTEMS (BY EVAN GREENBERG [29]) 29 FIG. II-28 – SIMPLE EA’S STEPS 32 FIG. II-29 – GENERAL EVOLUTIONARY ALGORITHM: I: INITIALIZATION, F(X): EVALUATION, ?: STOPPING CRITERION, SE: SELECTION, CR: CROSS-OVER, MU: MUTATION, RE: REPLACEMENT, X*: OPTIMUM. AUTHOR: JOHANN "NOJHAN" DRÉO 33 xix
  • 24. FIG. II-30 - DE OPTIMIZATION PERFORMANCE (PEDERSEN, M. [31]) ON SEVERAL DIFFERENT PROBLEMS USING DE/RAND/1/BIN ALGORITHM. PLOTS SHOW THE MEAN FITNESS ACHIEVED. OVER 50 OPTIMIZATION RUNS 34 FIG. II-31 – 3D ANAGLYPH AND ACTIVE STEREO VISUALIZATION ON A MULTITOUCH DI TABLE (MTT4ALL MULTITOUCH TABLE WAS BUILT BY THE AUTHOR OF THIS MASTER THESIS) 41 FIG. II-32 – INITIAL MTT4ALL SCALE MODEL (LEFT) AND FINAL MTT4ALL FUNCTIONAL PROTOTYPE IN USE , RUNNING FRAUNHOFER VIRTUALDESK APPLICATION, DEVELOPED BY THE AUTHOR OF THIS MASTER THESIS FOR FRAUNHOFER AUSTRIA (RIGHT) 42 FIG. II-33 – FRAUNHOFER, MULTITOUCH ARCHITECTURE VISUALIZATION – MESSE FRANKFURT GMBH (USING THE INSTANTREALITY FRAMEWORK) 43 FIG. II-34 – ANAGLYPH VISUALIZATION (CREATED BY THE AUTHOR OF THIS MASTER THESIS) 44 FIG. II-35 – ACTIVE STEREO VISUALIZATION (CREATED BY THE AUTHOR OF THIS MASTER THESIS) 44 FIG. II-36 – AN AUGMENTED REALITY SYSTEM DEVELOPED BY FRAUNHOFER (MONITOR + CAMERA + VIRTUAL REALITY SOFTWARE) 45 FIG. II-37 – DAVE, CGV AUSTRIA: IMAGES ARE PROJECTED ON THE BACK PROJECTION SIDE WALLS AND ON THE FLOOR FROM ABOVE, MMIRRORS ARE USED TO REDUCE THE SPACE NEEDED 45 FIG. II-38 – CGV AUSTRIA, THE DAVE, A 3D IMMERSIVE SYSTEM (EXPLORING THE NATIONAL LIBRARY OF VIENNA) 46 FIG. II-39 – THE SWEETHOME3D (MODELING APPLICATION) OUTPUT IS TRANSFERRED WITH A WEB SERVICE TO AN OPENSG CAVE APPLICATION WHICH LETS THE USER WALK THROUGH A 3D REPRESENTATION OF THE HOUSE PLAN 46 FIG. II-40 – HEYEWALL (HIGH RESOLUTION MULTITOUCH SCREEN) 47 FIG. II-41 – COMPUTER GENERATED HOLOGRAPHY. A COMPUTER CALCULATES A HOLOGRAPHIC FRINGE PATTERN FOR DISPLAY BY THE SPATIAL LIGHT MODULATOR (SLM), WHICH DIFFRACTS LASER LIGHT TO YIELD AN INTERACTIVE, TRUE 3D IMAGE 48 FIG. II-42 – TRADESHOW (AN HOLOGRAM SYSTEM) 49 FIG. II-43 – ALTHOUGH HE WAS IN MELBOURNE, TELSTRA'S CHIEF TECHNOLOGY OFFICER, HUGH BRADLOW (RIGHT), MAKES IS PRESENCE FELT AT A CONFERENCE IN ADELAIDE (PHOTO: TELSTRA) 49 FIG. II-44 – TOUCHABLE HOLOGRAPHY INTERACTION SYSTEM. AN AERIAL IMAGING SYSTEM, A NON-CONTACT TACTILE DISPLAY AND A WIIMOTE-BASED HAND-TRACKING SYSTEM ARE COMBINED. IN THIS FIGURE, THE ULTRASOUND IS RADIATED FROM ABOVE AND THE USER FEELS AS IF A RAIN DROP HITS HIS PALM 50 FIG. II-45 – 3D PHOTOREALISTIC RENDERING (CREATED BY HARCHI, AN ARCHITECTURE COMPANY BASED IN PORTUGAL) 51 FIG. II-46 – IPL/CDRSP FUTURE BUILDING (RENDERED IN AUTODESK MAYA 2011 BY THE AUTHOR OF THIS MASTER THESIS) 52 FIG. II-47 – MEDIUM QUALITY RENDERING OF A FACTORY INSTALLATION (CREATED IN DEEP EXPLORATION BY THE AUTHOR OF THIS THESIS) 52 FIG. II-48 – HIGH QUALITY REAL-TIME INTERACTIVE RENDERING (WWW.ICREATE3D.COM) 53 FIG. II-49 – MANIPULATION OF VR DATA, PROVIDED BY MEMPHIS [1] (TICIVIEW VR SYSTEM). IGI/FRAUNHOFER RESEARCH GROUP, 2007, SEOUL - SOUTH KOREA, (THE AUTHOR OF THIS MASTER THESIS WAS A MEMBER IN THE TEAM RESPONSIBLE FOR THE DEVELOPMENT OF THIS SYSTEM) 55 FIG. II-50 – ADVANCED INTERACTIVE VISUALIZATION (USING IVIEWER), STARTING FROM LEFT TO RIGHT: (A) SKYSCRAPPER; (B) AN APARTMENT INSIDE THE SKYSCRAPPER (WWW.ICREATE3D.COM) 55 xx
  • 25. FIG. II-51 – VIRTUAL FACTORY SIMULATION/SERIOUS GAME THAT WILL ALLOW A COMPANY TO GIVE TRAINING TO USERS, (THIS PROJECT WAS CREATED AT CDRSP BY THE AUTHOR OF THIS MASTER THESIS) 56 FIG. II-52 – SCREENSHOT OF THE TRICORDER DEVICE SHOWING THE FLOORPLAN OF A LAB, OVERLAYED WITH PLUG ICONS TO REPRESENT SOUND, LIGHT, CURRENT CONSUMPTION, MOTION AND VIBRATION. ALSO AVERAGE DATE FROM ALL SENSORS IS DISPLAYED [52] 58 FIG. II-53 – A PORTAL IN SECOND LIFE SHOWS SENSOR DATA OVER TIME 59 FIG. II-54 – A VIRTUAL DATAPOND IN THE VIRTUAL ATRIUM (LEFT) AND A REAL DATAPOND IN THE REAL MEDIA LAB ATRIUM (RIGHT) 60 FIG. II-55 – DIFFERENT REPRESENTATIONS OF A PERSON DETECTION IN 3D, STARTING FROM LEFT TO RIGHT: (A) BILLBOARD WITH A THIN COLORED SURROUNDING LINE, (B) ADDITIONAL GEOMETRY, (C, D) OVERLAY MARKER WHICH IS NOT OCCLUDED BY THE SCENE 60 FIG. II-56 – A MODEL WHICH IS A KINETIC PAVILION THAT REACTS ON WEATHER DATA 61 FIG. II-57 – PHASES OF THE REVERSE ENGINEERING PROCESS 63 FIG. II-58 – REVERSE ENGINEERING - CLASSIFICATION TECHNIQUES FOR 3D DATA ACQUISITION 64 FIG. II-59 – STEINBICHLER COMET 5 PHOTOGRAMMETRY 3D SCAN EQUIPMENT AT CDRSP REVERSE ENGINEERING LABORATORY 65 FIG. II-60 – A 3D SCAN OF A REAL GRAPHITE ELECTRODE USED IN MOLDS INDUSTRY. (3D SCAN AND ANALYSIS/INSPECTION DONE BY THE AUTHOR OF THIS MASTER THESIS, USING STEINBICHLER COMET 5 EQUIPMENT, AND STEINBICHLER COMET PLUS AND COMET INSPECT SOFTWARE) 65 FIG. II-61 – “SANTUÁRIO DO SENHOR DA PEDRA”, ÓBIDOS, PORTUGAL 66 FIG. II-62 – POINTS OBTAINED FOR THE “SANTUÁRIO DO SENHOR DA PEDRA” BUILDING 66 FIG. II-63 – 3D CLOUD OF POINTS FOR THE “SANTUÁRIO DO SENHOR DA PEDRA” BUILDING 66 FIG. II-64 - A) STL AND B) 3D MODEL 67 FIG. II-65 – SCALE MODEL OF “SANTUÁRIO DO SENHOR DA PEDRA” OBTAINED BY RAPID PROTOTYPING 67 FIG. II-66 – MAIN COMPONENTS OF A STEREO-LITHOGRAPHY MACHINE 69 FIG. II-67 – PROTOTYPES PRODUCED USING STEREO-LITHOGRAPHY 70 FIG. II-68 – SIMPLIFIED FDM PROCESS 71 FIG. II-69 – SCALE MODELS OBTAINED USING FDM 72 FIG. II-70 – SCHEME OF THE LOM PROCESS 73 FIG. II-71 – SCALE MODEL FOR OPORTO MUSIC HOUSE, PRODUCED USING LOM (PORTUGAL) 73 FIG. II-72 – SCHEME OF THE SELECTIVE LASER SINTERING (SLS) 74 FIG. II-73 - 3D PRINTING PROCESS 75 FIG. II-74 – COMPLETE SCALE MODEL OBTAINED USING THE 3D PRINTING PROCESS 75 FIG. II-75 – COMBINING SEVERAL TECHNOLOGIES/PROCESSES 76 FIG. II-76 - A BRUSH MADE IN A 3D PRINTER, USING TWO DIFFERENT MATERIALS, PRINTED SIMULTANEOUSLY INTO A SINGLE AND NOT ASSEMBLED FUNCTIONAL OBJECT (OBJECT INC.) 77 xxi
  • 26. FIG. II-77 – BIM VIRTUAL INFORMATION (VIRTUAL SIMULTANEOUS VISUALIZATION OF SIX DIFFERENT PHASES OF AN ONGOING BUILDING PROJECT, CREATED USING GRAPHISOFT ARCHICAD PLATFORM) 79 FIG. II-78 – VISION FOR AN AUTOMATED SYSTEM FOR AUTONOMOUS CONSTRUCTION OF BUILDINGS (BEHROKH KHOSHNEVIS) 80 FIG. II-79 – A REAL PROTOTYPE FOR AN AUTOMATED SYSTEM THAT WILL ALLOW THE CREATION OF BUILDINGS 80 FIG. III-1 – ENERGYPLUS SIMULATION ZONES 82 FIG. III-2 – DESIGNBUILDER AND ITS BUILDINGS ENERGY EFFICIENCY RATING 83 FIG. III-3 - WORKING IN ENERGYPLUS-MODE INSIDE ECOTECT, WHEN DEFINING OPERATIONAL SCHEDULES 84 FIG. III-4 - INTERNAL DAYLIGHT FACTORS SHOWN OVER A STANDARD WORKING PLANE 85 FIG. III-5 - OVERLAYING A SUN-PATH ON THE MODEL VIEW 85 FIG. III-6 - ANNUAL CUMULATIVE SOLAR RADIATION OVER THE EXTERNAL SURFACES 86 FIG. III-7 – COLOURED CONTOURS OF THERMAL CONFORT IN A CONFERENCE ROOM PREDICTED FOR A PARTICULAR VENTILATION SYSTEM DESIGN (ANSYS, INC. PROPRIETARY) 88 FIG. III-8 – ANSYS CFD MODELLING OF REGIONAL FLOW PATTERNS NEAR CAPE SHOPPING CENTRE (STEPHAN SCHMITT & THOMAS KINGSLEY; QFINSOFT, SA) 89 3 FIG. IV-1 – DIFFERENT POSSIBLE BUILDINGS FORMS, ALL WITH 1000M OF VOLUME, THIS CAN ALLOW THE COMPARISON OF RESULTS OBTAINED WITH DIFFERENT FORMS (PHD THESIS OF CHRISTINA LEMKE [88]) 92 FIG. IV-2 – DEFINITION OF AN ELEMENTARY VOLUME, ACCORDING TO DEPECKER, P., ET AL. [91] 93 FIG. IV-3 – FLOW FIELD AT A STREET INTERSECTION WITH A TALL BUILDING, ILLUSTRATING EXCHANGES BETWEEN THE STREETS AND ADDITIONAL MIXING PROCESSES DUE TO THE LARGE BUILDING 94 FIG. IV-4 – VIEW OF GREENHOUSE SHAPES IN E-W ORIENTATION 96 FIG. IV-5 – 3D MODELS THAT REPRESENT REAL WORLD FACTORIES 98 FIG. IV-6 - SELECTED BASIC FORMS INSPIRED IN REAL WORLD BUILDING SHAPES, STARTING FROM LEFT TO RIGHT: (A) CUBE, (B) CLASSIC AND (C) CYLINDER) CREATED USING EUCLIDES AND RENDERED USING DEEP EXPLORATION 98 FIG. IV-7 – TESTS FOR CREATING DIFFERENT 3D SHAPES (AND CONTROLLING ITS PARAMETERS) USING PARAMETRIC EQUATIONS 99 FIG. IV-8 – DEFINITION OF PARAMETRIC EQUATIONS IN MAPPLE 14 100 FIG. IV-9 - OPTIMIZED CODE GENERATION PRODUCED BY MAPPLE 14 100 FIG. IV-10 – PIECE OF JAVASCRIPT CODE TO GENERATE THE 3D CYLINDER SHAPE (CODE ADAPTED FROM PARAMETRIC EQUATIONS AND MAPPLE 14) 101 FIG. IV-11 – RESULTING JAVASCRIPT/EUCLIDES INTERFACE THAT ALLOWS THE CONTROL OF EACH SHAPE PARAMETER 101 3 FIG. IV-12 - 3D SHAPE GENERATION IN EUCLIDES (1.000 M OF VOLUME); FOLLOWED BY THERMAL ANALYSIS; AND PRESENTATION OF THE ANALYZED SHAPE. OTHER TYPES OF ANALYSIS CAN BE PERFORMED AS WELL. 105 3 FIG. IV-13 – ANOTHER 3D SHAPE GENERATION IN EUCLIDES (MAINTAINING 1.000 M ); FOLLOWED BY THERMAL ANALYSIS; AND PRESENTATION OF THE ANALYZED SHAPE 106 FIG. IV-14 – EXAMPLE LIST FOR MATERIALS THAT CAN BE USED INSIDE AUTODESK ECOTECT 109 FIG. IV-15 - OVERVIEW OF “OPTIMAL FORMS”, THE DEVELOPED AND PROPOSED GLOBAL OPTIMIZATION FRAMEWORK 110 xxii
  • 27. FIG. IV-16 – THE GLOBAL OPTIMIZATION FRAMEWORK RUNNING AUTONOMOUSLY (SIMULATION IN ECOTECT FOLLOWED BY 3D SHAPE GENERATION IN ORDER TO EVOLVE A POPULATION OF NEW BUILDINGS WITH DIFFERENT PARAMETERS USING THE DIFFERENTIAL EVOLUTION ALGORITHM) 111 FIG. IV-17 – THE ALGORITHM CHOOSES DIFFERENT INDIVIDUALS (BUILDINGS) FOR ANALYSIS, NOT STOPPING ON THE LOCAL MINIMA THAT WAS FOUND ALONG THE OPTIMIZATION PROCESS AND GIVING ROOM FOR JUMPING THOSE SAME LOCAL MINIMA 116 FIG. IV-18 – IN THIS RUN IT WAS GENERATED A COMPLETELY DIFFERENT INDIVIDUAL (BUILDING SHAPE), BUT THE ADMITTANCE VALUE WAS REALLY HIGH AND OTHER BETTER INDIVIDUALS WERE FOUND 117 FIG. IV-19 – A NEW OPTIMIZATION RUN, THIS TIME USING A DIFFERENT VALUE FOR CROSSOVER (0.6) 118 FIG. IV-20 – OTHER “TIGHTER” CONSTRAINS WERE CHOSEN, AND THE RESULTS WERE SLIGHTLY WORST THEN THE INITIAL ATTEMPTS (RUNS 1, 2 AND 3) WHERE A WIDER DOMAIN OF SEARCH WAS USED 119 FIG. IV-21 – THE OPTIMIZATION RUNS (1, 2 , 3 AND 4) ARE PLOTTED HERE SIMULTANEOUSLY 120 FIG. IV-22 – BY ALLOWING THE ALGORITHM TO GENERATE BUILDINGS THAT COULD USE A DIFFERENT ORIENTATION (LESS CONSTRAINED REGARDING THE ORIENTATION OF THE BUILDING), MORE EVALUATIONS WERE NEEDED, BUT A MUCH BETTER RESULT WAS OBTAINED 121 FIG. IV-23 – A NEW CONSECUTIVE OPTIMIZATION RUN (USING EXACTLY THE SAME VALUES) IN ORDER TO CHECK IF THE BEHAVIOR WAS CONSISTENT FROM RUN TO RUN 122 O O FIG. IV-24 – BY ALLOWING THE ALGORITHM TO SEARCH FOR SOLUTIONS IN AN ORIENTATION DOMAIN BETWEEN 0 AND 360 AND BECAUSE THE BUILDING DOES NOT HAVE DOORS OR WINDOWS YET, THE ALGORITHM FOUND A GOOD SOLUTION BY ORIENTING THE BUILDING ON A DIFFERENT DIRECTION (WHEN COMPARED TO RUNS 5 AND 5_1) 123 FIG. IV-25 – BY PLOTTING ALL THE INDIVIDUALS GENERATED BY THE GLOBAL OPTIMIZATION FRAMEWORK FOR RUNS (5, 5_1 AND 6) IT’S POSSIBLE TO CHECK THE CONSISTENCE OF RESULTS OBTAINED ON THESE MORE COMPLETE OPTIMIZATION RUNS 124 FIG. IV-26 – CASE STUDY 1 (FINAL CLASSIC BUILDING SHAPE) – DAILY AND ANNUAL SUN PATHS + SHADOWS AND DAYLIGHT TH LEVELS AT 14:00 PM, FOR THE 8 OF SEPTEMBER (BASED ON SPECIFIC CLIMATE/WEATHER FILES FOR COIMBRA, O 3 O PORTUGAL); WIDTH = 20 M; HEIGHT = 6 M; ROOF ANGLE = 60 ; VOLUME = 1000 M ; ORIENTATION: 122,10 125 FIG. IV-27 – WE CAN OBSERVE THAT AT EVALUATION 289 THE OPTIMIZATION FRAMEWORK HAD ALREADY ACHIEVED A VERY GOOD RESULT (COMPARED TO THE FINAL RESULT), BUT BECAUSE THE STOP CRITERION USED WAS, 1000 EVALUATIONS OR DX = 1.0, THE OPTIMIZATION CONTINUED UNTIL DX = 1.0 FOR FOUR CONSECUTIVE TIMES 128 FIG. IV-28 - CASE STUDY 2 (FINAL CUBE BUILDING SHAPE) – DAILY AND ANNUAL SUN PATHS + SHADOWS AND TOTAL TH RADIATION LEVELS AT 14:00 PM, FOR THE 8 OF SEPTEMBER (BASED ON SPECIFIC CLIMATE/WEATHER FILES FOR 3 O COIMBRA, PORTUGAL); WIDTH = 10 M; HEIGHT = 6 M; VOLUME = 1000 M ; ORIENTATION: 121,86 129 xxiii
  • 28. xxiv
  • 29. Tables List xxv
  • 30. xxvi
  • 31. TABLE III-1 - CHARACTERISTICS OF TWO DIFFERENT SIMULATION TOOLS [83] ................................................................. 87 TABLE IV-1 – ACTIVITY LEVEL IN AUTODESK ECOTECT ............................................................................................... 108 TABLE IV-2 - THIS TABLE SHOWS SEVERAL RUNS USING THE DEVELOPED FRAMEWORK IN ORDER TO OPTIMIZE THE DESIGN OF A “CLASSIC SHAPE FORM” BUILDING USING THE DIFFERENTIAL EVOLUTION (DE) ALGORITHM. IN THIS TABLE IS ALSO POSSIBLE TO SEE THE DIFFERENT CONSTRAINS AND PARAMETERS USED IN THE OPTIMIZATION PROCESS .................................... 114 TABLE IV-3 - RESULTS OBTAINED IN EACH OF THE OPTIMIZATION RUNS (ACCORDING TO TABLE IV-2).................................. 115 TABLE IV-4 – THIS TABLE SHOWS AN OPTIMIZATION RUN USING THE DEVELOPED FRAMEWORK IN ORDER TO OPTIMIZE THE DESIGN OF A “CUBE SHAPE FORM” BUILDING USING THE DIFFERENTIAL EVOLUTION (DE) ALGORITHM. IN THIS TABLE IS ALSO POSSIBLE TO SEE THE DIFFERENT CONSTRAINS AND PARAMETERS USED IN THE OPTIMIZATION PROCESS ....................... 127 TABLE IV-5 - RESULTS OBTAINED IN EACH OF THE OPTIMIZATION RUNS (ACCORDING TO TABLE IV-4).................................. 128 xxvii
  • 34. xxx
  • 35. A D ANM: Annealed Nelder and Mead strategy · 37 DAVE: Definitely Affordable Virtual Environment AR: Augmented Reality · 42, 44, 53, 54, 55 (Immersive VR System Developed by Fraunhofer) · ASA: Adaptive Simulated Annealing · 37 45, 46 ASHRAE: American Society of Heating, Refrigerating DDE: Dynamic Data Exchange · 87 and Air Conditioning Engineers · 87 DE: Differential Evolution · 4, 28, 32, 33, 34, 35, 36, 37, 39, 107, 112, 114, 127 B E BGA: Breeder Genetic Algorithm · 37 BIM: Building Information Modeling · 11, 19, 21, 78, 79 EA: Evolutionary Algorithm · 28, 32, 38 BLAST: Building Loads Analysis and System EASY: Evolutionary Algorithm with Soft Genetic Thermodynamics · 82, 83 Operators · 37 EDA: Estimation of Distribution Algorithm · 39 ES: Evolutionary Strategies · 37 C ESRI: Environmental Systems Research Institute · 10 ESTG: Superior School of Technology and Management CAD: Computer Aided Design · 8, 10, 14, 61, 63, 65, 78, ·2 79, 125, 133 Euclides: Fraunhofer JavaScript Procedural Modeler · CAM: Computer Aided Manufacturing · 14 2, 4, 5, 18, 81, 97, 98, 100, 101, 102, 103, 105, 106, CAVE: Cave Automatic Virtual Environment · 42, 45, 46 107, 109, 112 CC: Contour Crafting (Automated Construction System) · 80 CDRSP: Centre for Rapid and Sustainable Development F of the Product · vii, 2, 7, 41, 43, 52, 56, 65, 68 CEP: Complex Event Processing · 59 FAR: Floor Area Ratio · 12 CFD: Computational Fluid Dynamics · 88, 89, 133 FCT: Portuguese Foundation for Science and CGA: Computer Generated Architecture (shape Technology · 7 grammar) · 11 FDM: Fused Deposition Modeling · 68, 70, 71, 72 CGH: Computer Generated Holography · 47, 48, 49 CIBSE: Chartered Institution of Building Services G Engineers · 87, 104, 105, 141 CNC: Computer Numeric Control · 14 GA: Genetic Algorithm · 30 CPU: Central Processing Unit · 51, 52 GC: Bentley Microstation GenerativeComponents · 13 CR: Crossover · 34 GFA: Gross Floor Area · 12 CSG: Constructive Solid Geometry · 16 GIS: Geographic Information Systems · 10, 40 CT: Computed Tomography · 51 GML: Generative Modeling Language · 16, 17, 18 GPU: Graphics Processing Unit · 51, 52 xxxi
  • 36. H O HDE: Hybrid Differential Evolution · 39 OLED: Organic Light-Emitting Diode · 58 HSM: High Speed Machining · 68 HVAC: Heating, Ventilation and Air Conditioning · 82, P 83 PDM: Product Data Management · 8 I PLM: Product Lifecycle Management · 8 ICEO: IEEE Competition on Evolutionary Optimization · R 37 IEEE: Institute of Electrical and Electronics Engineers · Rhino: (a.k.a. Rhinoceros), it's a commercial NURBS- 37 based 3D modeling tool, developed by Robert IPL: Polytechnic Institute of Leiria · 2, 52 McNeel & Associates · 14; Rhinoceros (Robert McNeel & Associates) · 14, 15 L RICS: Royal Institution of Chartered Surveyors · 90, 140 LUA: Lightweight multi-paradigm programming S language designed as a scripting language with extensible semantics as a primary goal · 87, 102, SCG: Suicidator City Generator · 9 103, 141 SDE: Stochastic Differential Equations · 37 SLM: Spatial Light Modulator · 48 M SLS: Selective Laser Sintering process · 68, 69, 73, 74 STL: A file format native to the stereolithography CAD MTT: 2.3.1 Multitouch · 40 software · 67 MTT4ALL: Multitouch Table developed by the Author of this master thesis · 41, 42 T N Tabletops: Horizontal Interactive Displays · 40, 41 NP: Number of Elements in Each Generation · 34, 36, V 37 NURBS: Non-uniform rational basis spline · 14, 16 VEs: Virtual Environments · 53 VR: Virtual Reality · 42, 43, 53, 54, 55 VRML: Virtual Reality Markup Language · 43 xxxii
  • 38.
  • 39. I. Introduction 1 The Problematic The adjustment of architectural forms to local and specific solar radiation conditions is a fundamental study that must be always conducted by architects. When discussing energy consumption and solar power harness in buildings, important topics of discussion come into play, like the real relation between a building form and its energy behavior, or finding the right building shape for a specific location and weather conditions on an all year basis. Several studies were published so far, to try to answer and demonstrate these and other important questions. Form follows energy, but how exactly is this happening it‟s somehow difficult to demonstrate without having automated tools and models. One must try to manually analyze the energy dependence between form and volume. With this kind of studies, there is an attempt to simultaneously adapt a building form, in order to increase the potential areas for solar radiation “reception” and at the same time looks for ways to reduce the thermal loss (here the admittance method, well known by architects, it‟s useful), taking in account the need to design for specific locations and specific weather conditions. This research work aims, to examine the theoretical concepts associated to the problem of “Form Follows Energy”, pointed out, in studies done by some researchers. Also, the present study discusses emergent methods based on evolutionary algorithms and environmental simulation tools and it targets the development of new design methods that allow the construction of sustainable optimized buildings by using digital technologies, through the creation of an automated tool and an optimization framework that will allow the optimization of 3D shapes (buildings), taking in account the geo-location and specific weather conditions, throughout the run of automated simulations, making autonomous changes and optimizations utilizing evolutionary algorithms. 1
  • 40. For the creation of this work, a strong collaboration between the Polytechnic Institute of Leiria/Superior School of Technology and Management/Centre for Rapid and Sustainable Product Development (IPL/ESTG/CDRSP) and Fraunhofer Austria was established. The main research objectives of this work can be listed as follows: (i) To give an overview of state of the art technologies and techniques currently employed in the architecture field, regarding simulation and analysis, visualization, rendering, virtual interaction, rapid prototyping, reverse engineering and automated construction; (ii) To evaluate common shapes used in real world buildings, with focus on greenhouses forms as a practical example case study; (iii) To investigate, in order to obtain the necessary parametric equations (required to the use of computer graphics in the creation of procedural 3D models) for the several identified common building shapes. Also, to extract the essential parameters (height, width, length, orientation, roof angle…) of those fundamental shapes, in order to achieve a fully parametrical defined 3D model, this will allow the use of Euclides (JavaScript Procedural Modeler); (iv) To research on the possibility of having a programming integration with commonly used simulation packages and tools, to simulate how the different shapes of buildings have an influence in energy consumption throughout the life of these real buildings, with the final purpose of developing an automated tool capable of running automated simulations; 2
  • 41. (v) To make use of evolutionary algorithms in order to perform autonomous and automatic optimizations of 3D shapes based on automated simulations and the well-known method of admittance, always employing tools and methods which are widely accepted in the architecture field. The final goal is the creation of a global optimization framework for automatic generation of optimized 3D building forms, also taking in account the specific location weather conditions; (vi) To present and explain the importance of the results achieved with the developed global optimization framework, pointing out new directions in sustainable architecture design; Presently, this study is applied to architecture and sustainability. But it must be referred that this problematic of evolutionary architecture and simulation tools integration can be extended to other domains/fields, like the industrial or the medical field. They could also benefit from an autonomous generation of different 3D shapes, as well as a self-governing optimization of those same 3D forms. 3
  • 42. 1.1 Thesis Structure The thesis is divided into five chapters, which develops in accordance with the identified research objectives. This first chapter (Introduction) comprises an introduction that in addition to listing the key objectives, also briefly describes the context of the research. The contents of the remaining chapters are summarized as follows:  State of the Art Reviews the latest work developed around procedural modeling, visualization techniques, digital fabrication and reverse engineering. It also presents and describes a JavaScript Framework, named “Euclides”, utilized for the easy creation of 3D procedural and parametric shapes. This was the procedural framework used in this thesis for the creation of all the necessary parameterized 3D buildings forms. Lastly, a briefly explanation of how evolutionary algorithms work, is also given. Moreover, a specific evolutionary algorithm (Differential Evolution - DE) is described, as well as the reasons why this particularly algorithm was chosen in this thesis, for the development of an automated tool for 3D shapes (buildings forms) optimization. Other evolutionary algorithms are pointed out too, as plausible alternatives to be implemented within the optimization framework in future work, in order to tackle other problematic;  Simulation Tools in Architecture Gives an overview of different interests in simulation, in particular those related to the problematic of architecture and energy consumption in buildings. Some simulation tools/packages are presented, together with the reasons for selecting a particular tool to be used in the work presented here; 4
  • 43. A Global Optimization Framework Explains the work developed throughout this thesis, on the problematic of how buildings form affects the energy consumption on a daily basis throughout its entire life. Several methodologies used for choosing parameters to control a specific 3D shape as well as other tools used to deduce parametric equations and “mathematical code”, are also described with the objective to show how these 3D parametric models were generated using Euclides and JavaScript. Also, the general concept of the developed optimization framework is explained. A practical overview of the work is given, every framework component is presented in more detail and the achieved results are presented and clarified. Finally a case study is presented, where the problem of automatic optimization is extremely relevant and the results obtained are then presented and explained;  Conclusions and Future Work Provides an overall summary of the thesis and points out further progress paths and improvement options for the autonomous global optimization framework that was developed and presented in this thesis; 5
  • 44.
  • 45. II. State of the Art 2 Overview A review on the state of the art is presented, regarding current work focused on procedural, parametric and adaptive architecture modeling. A short description of evolutionary algorithms is given. Also, innovative methods of visualization and presentation of architecture projects are presented, as well as several techniques for rapid prototyping and reverse engineering. These methods and techniques are essential to capture 3D geometry, for achieving more complete results on any architecture project (e.g. production of scale models for simulation in wind tunnels, virtual simulation, building control…) and essential for presenting the achieved results to final customers (rendering, interactivity…), also, the author of this thesis was a research member of the CDRSP Research Centre and earned a scholarship, on the topic “Build-it-Green”, from the Portuguese Foundation for Science and Technology (FCT). This “Build-it-Green” topic is closely related to the architecture subject and some of the work that was developed at CDRSP by the author, was focused on these same areas and it was conducted throughout the realization of this master thesis. This state of the art review, aims to present a short explanation about each product, methodology or recent development. For getting more insightful details, the correspondent references should be further investigated. 7
  • 46. 2.1 Parametric and Procedural Modeling Parametric Computer Aided Design (CAD) modeling assumes, nowadays, an important role in the definition of 3D models. There are several active attempts to collect all the information about a product or about the different parts that compose a product. Information platforms like Product Data Management (PDM) or Product Lifecycle Management (PLM) [1], offer a way to gather the different distributed data that is vital for an efficient product management. However there is some “intelligent” information that must be captured with each part and product assembly, such as parametric information (e.g. width, height, volume, orientation, length, relations between parts, formulas …). Also semantic methods, using ontologies, try to present solutions for solving problems like the relationship between different, yet related 3D geometric information [2]. Procedural modeling can be viewed as the use of different techniques in computer graphics to create (generate) 3D models and textures from sets of rules. L-Systems, fractals, and generative modeling are procedural modeling techniques since they apply algorithms for producing scenes. The set of rules may either be embedded into the algorithm, configurable by parameters, or a set of rules that is completely separated from the evaluation engine. The output is then called procedural content, which can be used in computer games, films, be uploaded to the internet while requiring much less bandwidth, or the user may edit the content manually [3]. Procedural models often exhibit database amplification, meaning that large scenes can be generated from a much smaller amount of rules. If the employed algorithm produces the same output every time, the output needs not to be stored. Often, it is sufficient to start the algorithm with the same random seed to achieve the same result. Although all modeling techniques on a computer require algorithms to manage and store data at some point, procedural modeling focuses on creating a model from a rule set. Procedural modeling is often applied when it would be too cumbersome to create a 3D model using generic 3D modelers, or when more specialized tools are required, this is often the case for plants, architecture or landscapes [4]. 8
  • 47. 2.1.1 Plugins for existent 3D Software: Blender, 3D Studio Max and Others There are many plugins available on the internet for use within commonly used 3D modeling software, like Autodesk 3D Studio Max, Autodesk Maya or the open source modeling software Blender and many others that allow the automatic generation of terrain, buildings or even cities in a procedural way. These plugins permit the creation of 3D models according to specified rules and custom parameters specified by the user, they can also be customized through the use of scripting languages like, Python or MEL. Suicidator City Generator (SCG) is a wonderful example of such plugin for use inside Blender. It is a Python script for Blender or in other words it is a program written in the Python programming language that runs inside the Blender environment [5]. It‟s not the purpose of this work to explain in detail how these plugins perform, however they must be mentioned here as an existent and possible path for the creation of generative components in today‟s 3D modeling software packages. Fig. II-1 – Suicidator City Generator (SCG) for Blender 9
  • 48. 2.1.2 CityEngine City Engine (now acquired by ESRI) is one of the most successful and powerful examples for procedural modeling, it‟s a standalone software that provides a unique conceptual design and modeling solution for the efficient creation of 3D cities and buildings, for professional users in entertainment, architecture, urban planning, Geographic Information Systems (GIS) and general 3D content production [4]. CityEngine was also tested in this master thesis study. Fig. II-2 – Procedural/parametric examples created in CityEngine The key highlights of CityEngine include [6]:  GIS/CAD Data Support and OpenStreet Map Import CityEngine supports industry standard formats like, ESRI Shape file or DXF which allow the import/export of any geo-spatial/vector data such as parcels, building footprints with arbitrary attributes, or line data to create street networks. To copy real cities or efficiently create an urban environment for our design, it‟s possible to use data from OpenStreet Map. Geospatial data of real cities can also be downloaded and directly imported it into CityEngine; 10
  • 49. Dynamic City Layouts and Street Networks Patterns An intuitive toolset is provided to interactively design, edit and modify urban layouts consisting of (curved) streets, blocks and parcels. Street construction or block subdivision is controlled via parametric interfaces, giving immediate visual feedback; CityEngine offers unique street grow tools to quickly design and construct urban layouts. Street patterns such as, grid, organic or circular, are available and the topography of the terrain is taken into account;  Rule-based Modeling Core Procedural modeling based on Computer Generated Architecture rules (CGA shape grammar) offers unlimited possibilities to control mass, geometry assets, proportions, or texturing of buildings or streets on a city-wide scale. We can define our own rules using custom textures/models in the node- or text-based rule editor;  Facade Wizard and Parametric Modeling Interface One can quickly create rules out of an image or a textured mass model with this simple and easy-to-use visual facade authoring tool. The resulting facade rules are size-independent, contain level-of-detail and can be extended with e.g. detailed window asset. A convenient interface to interactively control specific street or building parameters such as the height or age (defined by the rules) is provided and with the live mode, parameter modifications invoke the automatic regeneration of the 3D model;  Map-Controlled City Modeling and Reporting (Building Information Modeling - BIM for Cities) Any parameter of the buildings and streets can be controlled globally via image maps (for example the building heights or the land use-mix); this allows for intuitive city modeling and quick changes on a city-wide scale. Furthermore, terrains can be imported, aligned, and exported. Customized rule-based reports can be generated to analyze the urban design e.g. automatically calculate quantities 11
  • 50. such as Gross Floor Area (GFA), Floor Area Ratio (FAR), etc. Reports are updated automatically and instantaneously and can be made for whole city parts;  Industry-Standard 3D Formats CityEngine supports Collada, Autodesk FBX, 3DS, Wavefront OBJ and e-on software's Vue, which allow for flawless 3D data exchange; FBX and Collada support asset instancing, multiple UV-sets, grouping and binary encoding; furthermore, scenes can also be exported to RenderMan RIB or Mental Ray MI format. Textures can be collected during (batch) export;  Python Allows streamlining repetitive or pipeline-specific tasks with the integrated Python scripting interface (e.g. write out arbitrary meta-data or instancing information for each building, import FBX cameras, etc...). CityEngine is also available for Windows (32/64 bits), Mac OSX (64 bits), and Linux (32/64 bits). Fig. II-3 – CityEngine IDE and the Rule Editor Capabilities 12
  • 51. 2.1.3 Bentley – MicroStation Extension: GenerativeComponents (GC) Designers have (since the dawn of times), wanted to innovate. Indeed, innovation is widely regarded as a trophy that awaits creative professionals who successfully explore endless design alternatives to ultimately arrive at the most efficient solution - a process that can be incredibly time consuming as each alternative is thoroughly modeled and assessed. Using the existing tools, a minor change to a design may require a major update to the model, thus restricting the number of design alternatives considered by the team due to time constraints. GenerativeComponents is an associative parametric modeling system used by architects and engineers to automate design processes and accelerate design iterations. As an innovation by MicroStation, GenerativeComponents extends proven technologies and delivers significant advantage to users as they rapidly explore a broad range of design alternatives. With a hybrid approach, designers who use GenerativeComponents can model geometry, capture relationships, and generate forms using scripts and/or direct manipulation for unrivalled creative flexibility [7]. Fig. II-4 – GenerativeComponents (GC) IDE, Bentley MicroStation This combination of accelerated iteration, flexible modeling, and automated process, means that a GenerativeComponents design can be highly efficient, benefiting from a combination of intuition and logic [7]. 13
  • 52. 2.1.4 Rhinoceros and Grasshopper Rhino (a.k.a. Rhinoceros) is a stand-alone, commercial NURBS-based 3D modeling tool, developed by Robert McNeel & Associates. The software is commonly used for industrial design, architecture, marine design, jewelry design, automotive design, CAD / CAM, rapid prototyping, reverse engineering as well as the multimedia and graphic design industries. Fig. II-5 – Rhinoceros is used in multiple fields, including architecture Rhino is specialized in free-form non-uniform rational B-spline (NURBS) modeling. Plug- ins developed by McNeel includes Flamingo (retrace rendering), Penguin (non- photorealistic rendering), Bongo and Brazil (advanced rendering). Over one hundred third- party plugins are available. There are also rendering plug-ins for Maxwell Render, V-ray and many other engines. Additional plugins for CAM and CNC milling are available as well, allowing for tool path generation directly in Rhino. Like many other modeling applications, Rhino also features a scripting language, based on the Visual Basic language and an SDK that allows reading and writing Rhino files directly. Rhino 3D gained its popularity in architectural design in part because of the Grasshopper plug-in for computational design. Many new avant-garde architects are using parametric modeling tools, like Grasshopper. Rhino's increasing popularity is based on its diversity, multi- disciplinary functions, low learning-curve, relatively low cost, and its ability to import and 14
  • 53. export over 30 file formats, which allows Rhino to act as a “converter” tool between programs in a design workflow. The combination between Rhino and Grasshopper is just perfect to create all kinds of parametric studies and developments on any field. The power offered by Rhino and Grasshopper is just amazing. There are also many other plugins available (rendering, math, physics, kinematics…). Fig. II-6 – Voronoi Examples, created using Rhinoceros and Grasshopper, by Atsuo Nakajima (Tokyo, Japan) For designers who are exploring new shapes using generative algorithms, Grasshopper is a graphical algorithm editor tightly integrated with Rhino‟s 3D modeling tools [8]. In Fig. II-7, by using Grasshopper, a building and its structural supports are generated and calculated using only two splines created initially in Rhino. Fig. II-7 - Parametric Strategies achieved using Rhino and Grasshopper. (Created by the author of this master thesis) Unlike RhinoScript, Grasshopper requires no knowledge of programming or scripting, but still allows designers to build form generators from the simple to the remarkable [8]. 15
  • 54. 2.1.5 Generative Modeling Language (GML) Traditionally, 3D objects and virtual worlds are defined by lists of geometric primitives: cubes and spheres in a Constructive Solid Geometry (CSG) tree, NURBS patches a set of implicit functions, a soup of triangles, or just a cloud of points. The term “generative modeling” describes a paradigm change in shape description, the generalization from objects to operations: A shape is described by a sequence of processing steps, rather than just the end result of applying operations. Shape design becomes rule design. This approach is very general and it can be applied to any shape representation that provides a set of generating functions, the “elementary shape operators”. Its effectiveness has been demonstrated, e.g., in the field of procedural mesh generation, with Euler operators as complete and closed set of generating functions for meshes, operating on the half-edge level [9]. Fig. II-8 - Creation of a simple house model using GML, the extrude operator is repeatedly applied to the ground polygon. To create the roof, the combined operator collapse-mid is applied to the faceCW and faceCCW edges of the edge returned by the extrude operation. Generative modeling, gains its efficiency through the possibility to create high-level shape operators from low-level shape operators. Any sequence of processing steps can be grouped together to create a new “combined operator”. It may use elementary operators, as well as other combined operators. Concrete values can easily be replaced by parameters which makes possible the separation of data from operations: the same processing sequence can be applied to different input data sets. Data can be used to produce different shapes by applying different combined operators, from, e.g., a library of domain-dependent modeling operators. This makes possible the creation of very complex objects from only a few high-level input parameters, such as, a style library [2]. 16
  • 55. Fig. II-9 – Parameterization/Configuration of a Chair with GML GML is a concrete implementation of the generative approach. Its main feature is that it is a full functional programming language that can nevertheless be efficiently used as a file format for low-level shape descriptions. Only 25 Kilobytes GML code of a Gothic window style library are sufficient to generate connected manifold control meshes for a variety of windows [10]. Fig. II-10 – Gothic Style Building generated with GML The GML comes with an integrated visualization engine. Thus, it can also be seen as a viewer with an integrated modeler that overcomes the usual separation of 3D modeling from interactive visualization. Curved parts are represented as subdivision surfaces that, within 1-2 seconds, unfold to seven million vertices after four steps of recursive refinement. The surface is adaptively displayed at interactive rates using optimized methods for culling and per-face per-frame multi-resolution rendering [11]. Fig. II-11 – Configuration of Different Wheel Rim Styles using GML 17
  • 56. 2.1.6 Euclides Framework and JavaScript Enabling an easy access to programming languages that are usually difficult on a direct approach will dramatically potentiate their use. GML [9] is such a language and can be described as being similar to Adobe‟s PostScript. A major drawback of all PostScript dialects is their unintuitive reverse Polish notation, which makes both - reading and writing - a burdensome task. According to Strobl, M., et al. [12] a language should offer a structured and intuitive syntax in order to increase efficiency and avoid frustration during the creation of code. To overcome this issue, Strobl, M., et al. [12] propose a new approach to translate JavaScript code to GML automatically. Within the last few years generative modeling techniques have gained attention especially in the context of cultural heritage. Because a generative model describes a rather ideal object and not a real one, generative techniques are a basis for object description and classification. This procedural knowledge differs from other kinds of knowledge, such as declarative knowledge, in a significant way. It can be applied to a task. This similarity to algorithms is reflected in the way generative models are designed: they are programmed. In order to make generative modeling accessible to cultural heritage experts, Schinko, C., et al. [13] created a generative modeling framework which accounts for their special needs. The result is a generative modeler called Euclides based on an easy-to-use scripting language (i.e. JavaScript). The generative model meets the demands on documentation standards and fulfills sustainability conditions and its integrated meta-modeler approach makes it independent from hardware, software and platforms. Fig. II-12 – Example of a 3D application created for this thesis using Euclides and JavaScript. It allows the control of several shape parameters on the “Classic Building form example”. 18
  • 57. 2.1.7 Autodesk Revit Architecture 2012 In the latest years, Autodesk made a strong effort in incorporating new technologies (e.g. multitouch…) and new amazing functionalities (physics, energy analysis, parametric design) in existing products like Autodesk Revit 2012 or Maya 2012, making a strong contribution for the development of really innovative products. Autodesk Revit Architecture can be used to create massing designs; explore design alternatives based on qualitative and quantitative feedback; and help address various environmental, constructability, and aesthetic concerns that can arise during project realization [14]. Fig. II-13 – Energy Consumption Study using Autodesk Revit In the early stages of a design, visualizing a concept in 3D enhances a designer‟s ability to communicate ideas. Analyzing these ideas yields the ability to predict and optimize the real-world performance of the built project. These attributes form a core value of the Building Information Modeling (BIM) process, which Revit Architecture software is purpose-built to support. 19
  • 58. In Autodesk Revit Architecture, users have access to a robust collection of easy-to-use modeling tools that facilitate design conceptualization, visualization, and communication. This software supports several new modeling operations, including adaptive, component- driven geometry, robust UV grid manipulation and increased schedule functionality through reporting parameters. In addition, Revit users on Autodesk Subscription can now access tools that enable them to better assess the impact of their early design decisions on energy consumption and carbon emissions without leaving the Revit environment. In order to clearly illustrate a complete workflow using the conceptual design and analysis tools and to address the new features introduced with the previous release [14]:  The Project requirements section outlines the criteria that will drive the building design;  The Parametric Massing Design section describes the steps taken to explore massing design alternatives informed by qualitative and quantitative feedback;  The Site and Environmental Analysis section addresses the impact of building mass and orientation on energy consumption and overshadowing;  The custom “Panelization” section uses the mass design options generated in the first section as the basis for informed panel‟s studies. Fig. II-14 – Conceptual Design in Autodesk Revit Architecture 20
  • 59. 2.1.8 Project Vasari and Project Nucleus Autodesk Project Vasari is an easy-to-use, expressive design tool for creating building concepts and it‟s build on the same technology as the Autodesk Revit platform. Project Vasari goes further, with integrated analysis for energy and carbon, providing design insight where the most important design decisions are made. And, when it‟s time to move the design to production, simply bring your Vasari design data into the Autodesk Revit platform for BIM, ensuring clear execution of design intent. Project Vasari is still under development and is primarily intended to reduce the building energy loads, not to replace the more detailed analysis tools. It is able to produce conceptual models using both geometric and parametric modeling functionality. The designs can be analyzed using the built-in energy modeling and analysis features. The tools depends on Green Building Studio (Autodesk‟s green building analysis web service) in many input energy related parameters [15]. Fig. II-15 – Sun Studies using project “Vasari” Project Vasari is focused on conceptual building design using both geometric and parametric modeling. It supports performance-based design via integrated energy modeling and analysis features. This new technology preview is now available as a free download and trial on Autodesk Labs. 21
  • 60. Project Nucleus integrates the Nucleus simulation engine from Autodesk Maya into Autodesk Revit Architecture and Project Vasari. It allows designers to experiment with "form-finding" in the conceptual design phase by simulating forces directly in Revit Architecture and Project Vasari (the latest technology preview of Project Vasari already includes the Project Nucleus functionality). Fig. II-16 - Panel Study using Revit and Vasari Project Nucleus can simulate a wide range of physical phenomena in real time, like wind, gravity, constraints, and collisions. These forces can help architects generate free-form shapes, many of which would be impossible to model by hand [16]. Fig. II-17 - Using Revit, Vasari and Nucleus Physics for a Panel Study, plus Analysis 22
  • 61. 2.1.9 Autodesk Adaptive Components Adaptive geometry can be sized and positioned in the context where it is used. When you designate under constrained geometry as adaptive, you specify the geometric elements allowed to change, while controlling the elements that you want to remain a fixed size or position [17]. “Adaptivity” is the functionality, within Inventor, that allows the size of a part/feature to be determined by setting a relationship to another part in an assembly. Basically, “adaptivity” is a special way to add constraints. These constraints differ from regular constraints in that they are driven from a separate file. This separate file can be an assembly file or another part within the assembly file. A good example of “adaptivity“, is constraining a shaft to a hole in another part. If set up correctly, when the size of the hole changes the diameter of the shaft updates as well. “Adaptivity” is normally used during the initial design phase of a model, when changes are made rapidly and many parts are affected. Fig. II-18 – Adaptive Components in Autodesk Revit 23
  • 62. Once a design is released, and parts become standard parts, available for use in other designs, “adaptivity” should be removed to eliminate the possibility of inadvertently changing a released design. Removing “adaptivity” also improves performance. Fig. II-19 – Adaptive Panel example in Autodesk Revit As with using any other constraint, forethought should be given to how a design may change before “adaptivity” is applied. If a part is not likely to change, it is better to apply normal (non-adaptive) constraints. “Adaptivity” should be used only when absolutely necessary [17]. Fig. II-20 – Another Adaptive Panel example, built using Adaptive Components 24
  • 63. 2.2 Evolutionary Architecture and the Use of Algorithms in Optimization of Problems The first references to this field of computation, Evolutionary Solvers or Genetic Algorithms [18], can be found in the early 60's when Lawrence J. Fogel published the revolutionary paper "On the Organization of Intellect" [19] which steered the first endeavors into evolutionary computing. The early 70's saw further ventures with important work produced by Ingo Rechenberg and John Henry Holland (and others) [20]. Evolutionary Computation didn't gain popularity beyond the programmer world until Richard Dawkins (one of my favorite authors) came out with the book, "The Blind Watchmaker" in 1986 [21], which was published with a small program that generated an apparently endless stream of body-plans called "Bio-morphs" based on human selection. Fig. II-21 – Image taken from the book “The Selfish Gene” by Richard Dawkins After the 80's, the dawn of the personal computer has made it possible for individuals without government funding to apply evolutionary principles to personal projects and making it a common jargon. The term "Evolutionary Computing" is very well commonly known at this time, but is still very much a programmer‟s tool (by programmers and for programmers) [18, 22]. 25
  • 64. The applications out there that apply evolutionary logic are either aimed at solving specific problems or they are generic libraries that allow other programmers to develop their own software [21]. Fig. II-22 – Several visions related to Evolutionary Architecture and Biomimetic One of the most important works ever published published about Evolutionary Architecture was the book of John Fraser, “An Evolutionary Architecture” [23], in the book introduction one can read: “…in this book the author investigates the fundamental form-generating processes in architecture, considering architecture as a form of artificial life, and proposing a genetic representation in a form of DNA-like code-script, which can then be subject to developmental and evolutionary processes in response to the user and the environment. The aim of an evolutionary architecture is to achieve in the built environment, the symbiotic behavior and metabolic balance found in the natural environment. To do so, it operates like an organism, in a direct analogy with the underlying design process of nature”. Fig. II-23 – Evolutionary examples taken from the book “An Evolutionary Architecture” by John Fraser [23] 26
  • 65. Also, Gordon Pask wrote on his foreword on this same book: “The book also proposes a fundamental change in practice… „The role of the architect here, I think, is not so much to design a building or city as to catalyze them: to act that they may evolve‟. Promising sustainable design methods are unquestionably emerging through the use of evolutionary computation and environmental simulation tools, as this is indeed an essential need in today‟s architecture world”. Fig. II-24 – Dynamic Geometry Computation, “Shanghai Tower - Geometry Generate and Rendering” (Michael Peng) Eugene Tsui on his work and book “Evolutionary Architecture: Nature as a Basis for Design” [24] and also Javier Senosiain, Michael Paulin, William McDonough, Renzo Piano and many others architects incorporate in their projects ecological and sustainable principles, but also integrate an understanding that constructions require “an holistic approach studying the form, materials and efficiency that Nature have becoming the infallible mentor in the creation of an comfortable and symbiotic world” [25]. Fig. II-25 – Ecological House of the Future (Eugene Tsui) 27
  • 66. 2.2.1 Differential Evolution (DE) Differential Evolution (DE) [26] has been very successful in solving the global continuous optimization problem [27]. It mainly uses the distance and direction information from the current population to guide its further search. The global optimization problem arises in almost every field of science, engineering or business, and an enormous amount of effort have been devoted to solving this problem. The major challenge of the global continuous optimization is that the problems to be optimized may have many local optima (Sun, J., et al. [27]). Fig. II-26 – The Global Optimization Problem (example in Matlab). Search of the highest mountain peak among a neighborhood of other high mountains peaks. Evolutionary Algorithms (EA‟s) are similar to the evolution process of a biological population which can adapt to the changing environments in order to find the optimum of the optimization problem by evolving a population of candidate solutions. Differential Evolution (DE) is one of the most successful EAs for the global continuous optimization problem. Several examples of problem solving using DE were already presented in the past, particularly those ones presented by the creators of the DE algorithm (Price, K. S., et al., [28]). 28
  • 67. Evan Greenberg [29] discusses in is master thesis a natural behavior called “Branching” that occurs in natural systems for functional reasons. The branching logic for each specific system is quite different due to environmental and mathematical factors. In the computation of branching systems, these mathematical factors can be incorporated easily into the coding of each system. Nevertheless, the environmental components deserve further consideration in the simulation of these natural systems. Through the engine of genetic algorithms based on evolutionary developmental theory, the specific logics observed and analyzed in branching patterns of river systems or trees, can be simulated and optimized in a digital environment. Fig. II-27 - Observation, Analysis and Computation of Branching Patterns in Natural Systems (by Evan Greenberg [29]) There are some biological terminologies which are used in evolutionary algorithm implementations, such as:  Individual: an autonomous piece characterized by a chromosome. In this case, one possible solution to the design problem;  Population: a group of individuals;  Population Size: the number of individuals in a population Gene (a functional block of DNA); 29
  • 68. Allele: A possible value of a gene;  Chromosome: Strings of DNA. In this case, a list of parameters;  Locus: The place of a gene in a chromosome. In evolutionary algorithms there are also three different types of operators: Selection, Crossover, and Mutation. After initializing the parameters, these three operators are iterated until the results satisfy the terminal criteria defined. Each step of the algorithms is explained as follows (Kawakita, G. [30]):  Initialization - In this step, some parameters including the population size, number of generations and so on are entered. After that, the initial input randomly generates genotype individuals of the first generation. Particularly, population size is significant in terms of the operations, the lengthier the chromosome length is, the bigger the population size is. Additionally, a bigger population size requires longer calculation time until convergence. However, small population sizes may result in premature and undesirable convergence;  Evaluation - Fitness scores are calculated for further selection of fitter chromosomes. One of the most important aspects in this step is the Fitness Function which calculates the fitness measurement of each individual. This operation is deeply related to the efficiency of the whole Genetic Algorithm (GA) flow; therefore, it needs to be determined carefully;  Selection - The fitter chromosomes in the population are basically selected for reproduction. As in biological evolution, the fitter chromosomes are more likely to be selected and reproduced in each generation. Meanwhile, lower fitness chromosomes are also possibly selected, but with a lower probability. This probabilistic selection depends on the selection method. There are several types of selection such as elite selection, roulette selection, tournament selection, etc. Each selection type has advantages and disadvantages. For instance, in elite selection, the 30
  • 69. fitter chromosomes are certainly selected in order; however, premature convergence is highly possible;  Crossover - Crossover roughly mimics the genetic operation of biological recombination between two chromosomes. The fitter chromosomes are chosen by the selection operator. However, it is not effective enough to evolve the population. The crossover operator encourages more variation by exchanging genes between two chromosomes;  Mutation - The mutation operator randomly flips or changes genes in a chromosome between alleles, generally with a very low probability. Chromosomes generated by the crossover operator are basically copies of the parent chromosomes. Therefore, premature convergence possibly occurs. Chromosomes that have been mutated help to avoid premature convergence. Generally speaking, the mutation rate should be 1/L, where L is the length of chromosome. Moreover, if the mutation rate is too big, the algorithm becomes similar to a random search;  Terminal Criterion - In this step, the conditions required to terminate the algorithm is evaluated. If the process is regarded as being completed, the fittest individual in the generation is outputted as one of the possible optimum solutions. The general conditions of convergence in evolutionary algorithms are as follows:  If the fittest score in the population satisfies the certain target – star gene;  If the average fitness score in the population satisfies the certain target – population improvement;  If the increase or decrease of fitness scores in the population becomes below a certain value – convergence;  If the number of generations becomes over the defined value – finite iteration. 31
  • 70. Fig. II-28 – Simple EA’s steps As it happens with every algorithm, there are several different variations of the differential algorithm, in order to classify the different variants, the notation: DE/x/y/z was introduced, where:  x specifies the vector to be mutated which can be “rand” (a randomly chosen population vector) or “best” (the vector of lowest cost from the current population);  y is the number of difference vectors used;  z denotes the crossover scheme. Example: “bin” (Crossover due to independent binomial experiments). Using this notation, the basic DE-strategy that is generally described can be written as: DE/rand/1/bin, but there are other variants, e.g. DE/best/2/bin. 32
  • 71. The DE algorithm work‟s (in general) as follows (Price, K. S., et al.[28]): 1. The DE algorithm maintains a population of N points in every generation, where each point is a potential solution and N is a control parameter; Then the algorithm evolves and improves the population iteratively: 2. In each generation, a new population is generated based on the current population; 3. To generate descendants for the new population, the algorithm extracts distance and direction information from the current population members and adds random deviation for achieving diversity; 4. If an offspring has a lower objective function value than a predetermined population member, it will replace this population member; 5. This evolution process continues until a stopping criterion is met (e.g., the current best objective function value is smaller than a given value or the number of generations is equal to a given maximum value). Fig. II-29 – General Evolutionary Algorithm: i: initialization, f(X): evaluation, ?: stopping criterion, Se: selection, Cr: cross-over, Mu: mutation, Re: replacement, X*: optimum. Author: Johann "nojhan" Dréo 33
  • 72. The optimization method known as Differential Evolution (DE) has several parameters that determine its behavior and efficacy in optimizing a given problem. The selection of good parameters for DE it is an important question that is discussed by Pedersen, M. [31], this paper gives a list of good possible choices of parameters values for various optimization scenarios with the intention to give an easy help when choosing the best values for achieving the best results, these interrelated parameters are: Crossover (CR), usually a good initial value for CR would be 0.9 or 1.0 to check if a quick solution is possible, Number of evaluations (Fitness Evaluations), Number of elements in each generation (NP), a good value for NP is between 5*D (D = Dimension of the problem) and 10*D, but NP must be at least 4 to ensure that DE will have enough mutually different vectors to which to work on, Differential Weight (F), a good initial value for F is usually 0.5, Number of variables of the problem (Problem Dimensions) and Size of the domain for each variable of the problem. Fig. II-30 - DE optimization performance (Pedersen, M. [31]) on several different problems using DE/rand/1/bin algorithm. Plots show the mean fitness achieved. Over 50 optimization runs 34
  • 73. In simple terms, optimization is the attempt to maximize a system‟s desirable properties while simultaneously minimizing its undesirable characteristics. What these properties are and how effectively they can be improved depends on the problem at hand (Price, K. S., et al. [32]). The optimization method known as Differential Evolution (or DE) was originally introduced by Storn and Price [32] and offers a way of optimizing a problem without using its gradient. This is particularly useful if the gradient is difficult or even impossible to derive (Pedersen, M. [31]). DE maintains a population of agents which are iteratively combined and updated using simple formula to form new agents. The general purpose optimization method known as DE has a number of parameters that determine its behavior and efficacy in optimizing a given problem, these parameters must be chosen accordingly [31, 32]. Small changes to the DE implementation can cause dramatic changes in the behavioral parameters that cause good optimization performance. The parameters given by Pedersen [31] have been tuned for the DE/rand/1/bin algorithm. If other DE implementation is chosen, different parameters values must be selected. According to Storn and Price [26], the DE algorithm (DE/rand/1/bin) was demonstrated to converge faster and with more certainty than many other acclaimed global optimization methods. DE also requires fewer control variables, it‟s robust, easier to use and fits itself on parallel computation scenarios. Storn and Price [26] compared and tested DE against other optimization methods in solving minimization problems. The DE algorithm uses simultaneous search vectors in order to help escape local minima (not using only a usual greedy criterion, where a new parameter vector is accepted if an only if it reduces the value of the cost function). Also DE was compared to Simulated Annealing [33] (annealing relaxes the greedy criterion by occasionally permitting an uphill move, allowing a parameter vector to climb out of a local minimum, however in a long run this method also leads to a greedy criterion). 35