Interactive Structural Analysis

Odysseas Georgiou University of Bath

London, UK o.georgiou@bath.ac.uk

Abstract This paper re-approaches structural engineering through an interactive perspective by introducing a series of tools that concatenate parametric design with structural analysis, thus achieving interoperability between form and its structural performance. Parametric Design is linked to Structural Analysis using computer programming to establish a common interactive framework that leads to performance based designs that respond to structural constrains and conditions in an interactive manner. A series of examples illustrate the synergy between form and structure by interactively modelling, analysing and visualizing its response.

1. Introduction The digital age has introduced new ways of approaching architectural design. Modern digital tools (CAD), allow geometric freedom that in turn enables designers to overcome traditional design boundaries and transform almost any form into a constructible building. The development of complex forms in architecture led to the transition from the traditional, static design approach, to a “dynamic” process [1], [3] and forced computers to become “design generators” [2] and CAD tools to move away from their conventional drafting role. The development of generative design methods, or commonly parametric design, enables the designer to describe mathematically complex forms and explore new ones while understanding their conceptual functions.

Since building forms are becoming more complex, the engineering input is becoming essential from the concept stages and the role of the engineer in a design collaboration is enhanced. Despite that, there seems to be a gap between the ways the two disciplines interact. While a concept form can be shaped, understood and altered quickly by the architect, an engineer cannot assess the data given in analogous speed. For a given free form shape, developed with ease inside a Parametric CAD environment, extensive analysis needs to take place on the engineering side for it to be structurally assessed. Most of the times, the full behaviour is not fully understood due to the time chasm between modelling, analysis and assessment and just static performance feedback is returned for the shape analysed. Structural constraints, such as performance of structural members or materials and boundary conditions for example, are not controlled equally by Structural Analysis software as opposed to the way geometrical constraints are treated by Parametric Design software. Conceptual structural engineering of complex shapes cannot follow the rapid and constantly shifting architectural object that it is summoned to support. The digital design process needs to be enhanced and be able to stimulate solutions which respond to structural performance and conditions. There are a few examples that implement the notion of structural performance models applied on specific projects as Phaeno Science Center [4] and Lansdowne Road Stadium [5]. A more global example of performance based design of roof trusses can be found in Ref [2]. Moreover, there are cases that implement other types of data as lighting or thermal information to generate performance based models [6]. This paper illustrates a synergy between the architectural shape and its structural performance through an interactive link between a Parametric Design tool, Grasshopper 3d [7], and a Structural Analysis Software, Autodesk’s Robot Structural Analysis [8] and demonstrates how performance can be incorporated into the design process in order to firstly generate an analytical model of the given shape and secondly iterate between a set of efficient structural solutions that apply to that shape. 2. Methodology To illustrate an overview of a performance based structural design, a series of definitions were assembled by linking together McNeal’s Grasshopper 3D and Autodesk’s Robot Structural Analysis [19] using C# programming language.

Grasshopper 3d is a plugin built in .NET framework to access McNeal’s core software, Rhinoceros 3d in order to control and manipulate geometry in a generative manner. The functionality of Grasshopper 3D (GH) can be extended by writing code in C# or VB .NET programming language to create custom components. In parallel, Robot Structural Analysis (RSA) allows the interaction with other software and the use of its Calculation Engine through an Application Programming Interface (API). At first, a distinction should be made between the way that CAD software and FEM software understand and control geometry: when modelling a structure, a set of additional parameters should be taken into account in order to facilitate the process of analysing its performance. These parameters are acknowledged as structural modelling entities and include nodes, bars and panels and match real building elements as foundations, beams, slabs and so on. Some of these entities also exist in the pure geometric model, sometimes with different naming, and in fact determine a similar notion. For example, two points can describe a line and in the same way, two connected nodes describe a structural bar. What makes a significant difference in the representation of a structural model is the need to attach structural attributes to each of its elements, while the geometric model can be purely described by its topology. In addition, the numbering of each element and its global orientation in relation to its local axis definition are crucial points in the definition of a structural model (Figure 1). A simple bar element for example, is labelled with a number and is defined by its two interconnected nodes i and j, each one numbered individually in the global system. The accurate definition of nodes is important in structural analysis since that is where all the calculations occur.

Fig 1

Thus, eaccordinstructurand are define aa geomcontrollcalled Ra set ofperformmatch thactual sRobotNand z RobotNfor crea 2.1. Ex

The fodiamenschange

1- Local Coor

each node ngly to be reral bar, the lin

used to defia structural b

metric line toled using an Robot Objectf data, defin

med which arhe operationssoftware envi

Node interfaceglobal coo

NodeServer inating a new n

ample 1: Pa

ollowing exasional definiof a paramet

rdinate systemStructural A

that definesead by RSA.ne’s start andine two new bar since thero a bar. Eaappropriate i

t Model [9]. Aned as attribure called funs that a user ironment. Ae, which amoordinates. nterface, whiode.

arametric Tr

ample showition of a Trter inside Gra

m and sign ConAnalysis v.18 U

s geometry . For exampld end point cnodes for R

re does not each structurainterface, whAn interface utes or memnctions. Thefollows to m

A node for exong others in Each nodich for exam

russ

ws the proceruss where iasshopper 3d

nvention, ImagUser’s Manual

in GH nele in order tocoordinates a

RSA. Those nexist an implial member inhich is includis a software

mbers, and opese interface

model and anaxample can b

ncludes three de can be mple includes

edure followits performand’s visual env

ge courtesy of l

eds to be o translate a

are extracted nodes are theicit way of trn RSA is dded in the APe structure coperations thas and their alyse a structbe representereal numbermanaged u

s the Create

wed to modnce respondsvironment.

f Robot

modified a line to a

from GH en used to ranslating efined or

PI library, omprising at can be functions

ture in the ed by the rs for x, y using the e function

del a 2-s to each

The truattributeattributecoordinx, y anRobotNinterfacvalue anform strsequencby the Rstart and The GHconsequAt this propertiRSA’s define tstructurcalculatresults cresults type of shown ifrom thenable s

uss’s geomees could bees included

nates of the pnd z progra

NodeServer ine, each new nd its x,y,z ructural bars ce used in deRobotBarSerd end node n

H geometry iuently geome

stage, strucies and loadsrelevant intethem using tral model intion engine ican be generare returned

f the results ain Figures 2

he analysis. structural ana

Fig 2- Visu

etry was firse controlled the truss’ ov

points compriamming listsnterface for structural n

coordinates. using the Ro

efining the grver interfac

number value

is then fully etrical param

ctural constras need to be erfaces and mthe software

ncludes all tnterface (Rorated and retu

d by utilizingand the refer

2 and 3, nodaThese resul

alysis visuali

ualisation of d

st defined pin an inte

verall span, ising its geoms and were

structural nnode is create

Each subseobotBarServeeometric moe are each b

es.

associated wmeters implyaints such asincluded in

members in e’s Graphicalthe informat

obotCalcEngiurned inside g the approprence to the al deflectionlts can now isation in Rhi

deflection und

parametricalleractive mandepth and nmetric conne

then used nodes to beed by inputtiequent node er following odel. Relativebar’s global n

with RSA geoy structural ms supports, bthe model. Ta similar wal User Interftion needed ine) can be cthe GH cus

priate interfacstructural en

n and bar forbe used as pinoceros 3D

er self-weight

ly so that anner. The g

number of baectivity were

in sequencee created. Uing its globais then con

the same conely, the inputnumber valu

ometry and pmodelling pabar sections,This is done ay, as the usface (GUI).

to be analycalled and thtom componce, dependinntity. For exrce data wasparametric oenvironment

t in Rhino3D

all of its geometric ays. The stored in

e by the Using the al number nnected to nnectivity ts needed

ue and its

properties, arameters. , material by using

ser would When a ysed, the

he desired nent. The ng on the xample as s returned outputs to t

Fig 3- V

Furthermmodel, additionRhinocetruss’ toIn a simnumeric

Fig 4

A wide custom given btruss tyselectio

Visualisation o

more, as a mparametric c

n to the algoeros‘s enviroop chord whmilar mannercal slider in r

4 - Any arbitraposition can

range of parcomponent

by selecting ypes (Brownn from RSA

of the distribu

means to increcontrol to aorithm enabonment to beich can be cr, the truss’ respect to the

ary load case cbe changed us

rameters is pofor the trusmultiple loa

n truss, Pratsteel section

ution of forces Rhino3D

ease the functdditional strles any line

e used to appontrolled in support poin

e original edg

curve can be dsing a numeric

ossible to be s and in the

ad cases or ltt truss and n library and

in each eleme

tionality of thructural variae of load or ply a distribureal time by

nts can be mge supports (s

drawn in Rhinocal slider in G

added and coe current conload case co

Vierendeel truss length (

ent under self-

he performanables was ad

any curve uted load ca

y curve contrmoved by usisee Figure 4)

o 3D and the s

Grasshopper 3D

ontrolled witntext, such combinations,

truss), sect(see Figure 5

weight in

nce based dded. An drawn in

ase on the rol points. ing a GH ).

support D

thin GH’s control is different

tion type 5).

 

2.2. Ex

As the design mmodel agrillageGrasshostructurrespond

Fig 5 - Defi

ample 2: 3D

previous examodel, it did among the tw is linked to

opper 3d. Byre where itding to the ch

ining the truss

D Grillage

ample illustrnot show th

wo disciplineo a free-form

y this means,ts performahange of geom

Fig 6

’ parameters in

rated the gene possibility es. In this exm surface th, any compleance attributmetry.

6 - Free-form s

n Grasshoppe

neration of a for a commo

xample a 3-dhat is modelex architecturtes are vis

surface

r’s sketchpad.

parametric only shared gdimensional lled parametral skin can sualised inte

.

structural geometric structural

trically in acquire a eractively

In this cselectiosoftwareaccordin

The edgspacingthe surfthe grillto genenodal losection a genercomponStructurtasks (avisualisthe analnodes adistribu

case the paran and is thene. These rng to the load

ges of the g bays genera

face accordinlage. The seterate the 3-doading to be types and marative mannenents. It is thral Analysis as this toolse the currenlytical modelnd the constr

ution along th

ametric surfan transformeregions alsod-cases set b

Fig 7

generated regating structurng to a contrt of nodes aredimensional g

applied to taterial propeer by the

he user’s chosoftware, in

s’ capabilitient results in l in RSA inclrain conditio

he bars of the

ace is split ined to structuro form the y the user (F

- Division in

gions are subral nodes. Throllable parame then connegrillage. Thethe structuralerties are also

user, either ice to compun order to pes are limiteGrasshopperluding the pons (Figure 8)

e grillage in G

nto regions acral data to beloading stri

Figure 7).

regions

bdivided intohe nodes aremeter which ected with stre surface loal nodes. The o parameters

using numute the analyperform mored to concepr. The followoint loads ass) and the visuGrasshopper’

ccording to te read by theips of the

o the desiredthen offset ndefines the

ructural bars ading is tran constrain cothat can be derical slider

ytical model ire rigorous aptual analysiwing images signed on to ualisation of ’s envorment

the user’s e analysis structure,

d grillage normal to height of in space,

nslated to onditions, defined in rs or text inside the analytical is), or to illustrate

structural the stress

t.

 

Fig 8

Fig 9-

- Analytical m

- Visualisation

model in Robo

n of stress ratio

ot Structural A

os in Grasshop

Analysis

pper 3d.

2.3. Ex

A moretranslatis able alternatanticipaanalysewith indesigne

For the a Finitassessmneeds toused toparametshape mconnect

ample 3: Int

e complex ting any fre to withstante supportinated that sed and assesnstant resulter.

Fig

surface to bete Element

ment. Therefoo be convert

o mesh any trically in Gmore a unifotivity matrix

teractive An

case is invee-form surfnd the exteng system asuch modelsssed and thut feedback

10 – Free-for

e analysed asmesh with

ore, as a firted into a trisurface thatH. The mesh

orm discretisof vertices,

nalysis of a f

vestigated inface (Figureernal loadinas was the s demand aus being abwould gen

rm surface con

s a shell struch analytical st step to thiangular mest is either dh settings arsation of theedges and fa

free form sur

n this sectioe 10) to a strng condition

case of thea larger effole to contro

nerate an ef

ntrol-point han

cture it needsinformation

he process, tsh. A standadrawn in Rhe controlled

e surface. GHaces. This inf

rface

on, which rructural shens itself wie 2nd examport to be geol geometry fficient tool

 ndling

s to be transfn for its the free-formard GH comhino 3D or

by the compH’s mesh ouformation ne

relates to ell which ithout an ple. It is enerated, changes

l for the

formed to structural

m surface mponent is

described ponent to utput is a eeds to be

decomposed in elements that can be read by the structural analysis component. The mesh vertices are translated in a list of points in 3D space, which are then translated in structural nodes inside the analysis component. Up to this point, no relation between the nodes exists only an unreferenced point cloud is drawn in RSA. The connectivity of each face is then used to create arrays of nodes by selecting the three points that comprise each face out of the list. Finite elements can then be created by utilizing the arrays created using the FiniteElems interface in RSA. Structural properties need to then be applied, which are treated in a way that allows them to be controlled parametrically by the user through the GH graphical interface. The properties that are essential for the analytical process are: the type of the elements, which are set to planar 3-node elements with constant thickness, the material that either is pre-set from RSA’s library or can manually be defined by its properties and the element’s thickness. The model’s supports are initialised to the edges by selecting all the nodes that lie on the mesh’s outline. The border of the mesh is isolated by utilizing the topological information of the mesh by accessing Rhino.NET SDK [10]. A routine is formed to run through all the edges that form the mesh’s topology and select the ones that are “naked”, which means that they are only connected to one face. The vertices that belong to those edges are output to the analysis component and are selected as nodes that contain support information. The option of adding supports on points that did not necessarily coincided with the mesh vertices was also implemented by subdividing the mesh around the area to be supported. The topological information is lost when the mesh is translated to FE for the structural modelling and therefore the index of each support vertex needs to be found from the list of nodes that are already drawn in the model. After the structural information database is complete, the calculation interface can be called and results can be output for each finite element. For the purpose of graphical representation of the FEA results, each face was translated into a triangulated surface while keeping the topological information of the mesh. Consequently, the analysis results are mapped with each surface facet and their values can be tagged accordingly or coloured using a GH

Gradienimpact surface

Fig

3. ConAn Intethroughsoftwaredesign, problemdesign analyse a real-tisolutiondesign complex

nt componenof change oin Rhino’s V

g 11 - Graphic

nclusions gration betw

h the synerge. This coma function w

ms. As mentioare readily and manipul

ime manner. ns. The invois essential

x forms. Thi

nt. This resuon the geomViewport (see

cal representat

ween the formgy between mbination enwhich currentoned in the iavailable to

late models q This proces

olvement of in modern

s offers a qu

ults to the metry or the c

e Figure 11).

tion of surface

m and its struParametric

nabled an intly is sparingintroduction, o architects. quickly, whilss often leada structuralarchitecture

ick insight to

interactive rconstraints o

e stress ratios i

uctural perforDesign and

nteractive apgly applied fo tools that aiThese allow

le being ableds to architecl engineer ine and partico the structur

representatioof the initial

in Grasshoppe

rmance was id Structural pproach to or solving enid fast and asw them to

e to visualise cturally elegan the early cularly in reral performan

on of the l NURBS

er 3D

illustrated Analysis

structural ngineering ssociative generate, results in

ant design stages of egards to nce of the

form and aids increasing efficiency in the whole design process. However, without the use of compatible tools, an engineer not only fails to support architectural design but also moves away from efficient solutions. The generation of structurally responsive parametric models enables the iteration between ranges of efficient solutions while being able to react to each change of the architectural geometry rapidly. This paper demonstrated the application of the method through three examples: a simply supported truss, a 3D grillage and a shell structure of a free-form surface. The possibility for all three models to adapt to multiple variations of shape and the continuous user interaction was persistent in all cases. This also illustrates the methods’ applicability to a range of architectural and structural engineering problems.

3.1. Future Work

Future extensions to the performance based parametric components include:

- The initialisation of Robot Structural Analysis member design engine would allow designing and checking structural members while ensuring their structural efficiency.

- The implementation of improved subdivision algorithms for the generation of mesh emitters around the support points to improve the distribution of stresses related to the interactive analysis of a shell structures.

- Including a greater range of structural elements in a single model would allow the generation of performance models which contain largest amounts of assessable information which would enable a more holistic approach to the problem.

References

1. Aish, R.,Woodbury,R., Multi-Level Interaction in Parametric Design. Bentley Systems Incorporated.

2. Kristina Shea, Robert Aish & Marina Gourtovaia. (2005).Towards integrated performance-driven generative design tools. Automation in Construction. March 2005, 14(2).pp. 253-264.

3. Papalexopoulos D., Stavridou A., Papadopoulos D., 2007, Conceptual determination of parametric attributes of architectural construction elements and materials (free translation) . National Technical University of Athens, Athens.

4. Littlefield D., ed., 2008. Space Craft-Developments in Architectural Computing- Geometry form and complexity. London: RIBA Publishing.

5. Shepherd, P., Hudson, R., Parametric Definition of the Lansdowne Road Stadium. International Association of Shell and Spatial Structures. 2007, Venice, Italy.

6. Caldas, L., GENE_ARCH: An Evolution-Based Generative Design System for Sustainable Architecture. Intelligent Computing in Engineering and Architecture. Lecture Notes in Computer Science, 2006, Volume 4200/2006, 109-11.

7. Grasshopper- Generative modelling for Rhino [online]. Available from : http://www.grasshopper3D.com/ [Accessed 22 Dec 2010].

8. Autodesk- Autodesk Robot Structural Analysis Professional[online]. Availablefrom:http://usa.autodesk.com/adsk/servlet/pc/index?siteID=123 112&id=11818169 [Accessed on 20 Dec 2010].

9. Autodesk Inc., 2009. Autodesk Robot Structural Analysis 2010 – Robot Object Model.

10. Developer: dotnetplugins- McNeel Wiki [online] Availble from: http://wiki.mcneel.com/developer/dotnetplugins [Accessed 20 July 2010].