integrative design computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf ·...

10
72 INTEGRATION THROUGH COMPUTATION ACADIA 2011 _PROCEEDINGS In contrast to most other building materials, wood is a naturally grown biological tissue. Today, the organic nature of wood is recognized as a major advantage. Wood is one of the very few naturally renewable, fully recyclable, extremely energy efficient and CO2-positive construction materials. On the other hand, compared to industrially produced, isotropic materials, the inherent heterogeneity and differentiated material makeup of wood’s anatomic structure is still considered problematic by architects and engineers alike. This is due to the fact that, even today, most design tools employed in architecture are still incapable of integrating and thus instrumentalizing the material properties and related complex behavior of wood. The research presented in this paper focuses on the development of a computational design approach that is based on the integration of material properties and characteristics. Understanding wood as a natural composite system of cellulose fibers embedded in a lignin and hemicelluloses matrix characterized by relatively high strain at failure, that is high load-bearing capacity with relatively low stiffness, the particular focus of this paper is the investigation of how the bending behavior of wood can become a generative design driver in such computational processes. In combination with the additional integration of the possibilities and constraints of robotic manufacturing processes, this enables the design and production of truly material- specific and highly performative wood architecture. The paper will provide a detailed explanation of such an integrative approach to design computation and the related methods and techniques. This is complemented by the description of three specific research projects, which were conducted as part of the overall research and all resulted in full scale prototype structures. The research projects demonstrate different approaches to the computational design integration of material behavior and robotic manufacturing constraints. Based on a solution space defined by the material itself, this enables novel ways of computationally deriving both material-specific gestalt and performative capacity of one of the oldest construction materials we have. Integrative Design Computation INTEGRATING MATERIAL BEHAVIOUR AND ROBOTIC MANUFACTURING PROCESSES IN COMPUTATIONAL DESIGN FOR PERFORMATIVE WOOD CONSTRUCTIONS Achim Menges Stuttgart University Harvard University ABSTRACT

Upload: others

Post on 18-Jul-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Integrative Design Computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf · 2011-09-13 · Integrative Design Computation integrating material behaviour and robotic

72

integration through computationacadia 2011 _proceedings

In contrast to most other building materials, wood is a naturally grown biological tissue. Today,

the organic nature of wood is recognized as a major advantage. Wood is one of the very few

naturally renewable, fully recyclable, extremely energy efficient and CO2-positive construction

materials. On the other hand, compared to industrially produced, isotropic materials, the

inherent heterogeneity and differentiated material makeup of wood’s anatomic structure is still

considered problematic by architects and engineers alike. This is due to the fact that, even

today, most design tools employed in architecture are still incapable of integrating and thus

instrumentalizing the material properties and related complex behavior of wood.

The research presented in this paper focuses on the development of a computational

design approach that is based on the integration of material properties and characteristics.

Understanding wood as a natural composite system of cellulose fibers embedded in a lignin and

hemicelluloses matrix characterized by relatively high strain at failure, that is high load-bearing

capacity with relatively low stiffness, the particular focus of this paper is the investigation of how

the bending behavior of wood can become a generative design driver in such computational

processes. In combination with the additional integration of the possibilities and constraints

of robotic manufacturing processes, this enables the design and production of truly material-

specific and highly performative wood architecture. The paper will provide a detailed explanation

of such an integrative approach to design computation and the related methods and techniques.

This is complemented by the description of three specific research projects, which were

conducted as part of the overall research and all resulted in full scale prototype structures. The

research projects demonstrate different approaches to the computational design integration of

material behavior and robotic manufacturing constraints. Based on a solution space defined by

the material itself, this enables novel ways of computationally deriving both material-specific

gestalt and performative capacity of one of the oldest construction materials we have.

Integrative Design Computationintegrating material behaviour and robotic manufacturing processes in computational design for performative wood constructions

Ach im Menges

Stu t t ga r t Un i ve r s i t y

Ha r va rd Un i ve r s i t y

ABSTRACT

Page 2: Integrative Design Computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf · 2011-09-13 · Integrative Design Computation integrating material behaviour and robotic

73

Figure 1a. Madan Houses (Oliver 2007)

Figure 1b. Hooke Park (Nerdinger 2005)

Figure 1c. Hooke Park

computation, formation and materiality

Fig. 1a

Fig. 1b

Fig. 1c

1 Introduct ion: Integrat ive Design Computat ion

In arch i tecture, computat iona l des ign processes have been invest igated in numerous

ways over the last 50 years. In para l le l to the deve lopment of Computer A ided Design,

which is character ized by the t ransfer of long-estab l ished, representat iona l des ign

techniques in to the d ig i ta l rea lm, research in to generat ive Computat iona l Des ign has

been conducted for many decades. In the 1960s des ign programs such as GRASP

(generat ion of random access s i te p lans) deve loped by Er ic Te icholz at the Harvard

Laboratory for Computer Graphics, exp lored ways of generat ing rather than drawing

des ign so lut ions and eva luat ing the i r per formance (Howard 1998). Unt i l recent ly,

par t icu lar ly dur ing the last two decades, a wide range of generat ive computat iona l

des ign techniques have been both researched in academia and tested in pract ice

(Mark, Gross and Goldschmidt 2008). Rather than determin ing form, as is emblemat ic

for CAD, most of them share a conceptua l izat ion of des ign as a process of der iv ing

arch i tectura l form through the generat ion and a lgor i thmic process ing of in format ion

(Terz id is 2006). Thus form is understood as resu l t ing f rom the in teract ion of a

def ined system and externa l data, and d isp lay ing a par t icu lar per format ive capaci ty

when eva luated aga inst def ined des ign cr i ter ia . In th is way, form, in format ion, form

generat ion and per formance are inherent ly re la ted, and the degree of in tegrat ion

main ly depends on the number of system parameters and des ign cr i ter ia embedded

and eva luated in the computat iona l process. However, i t is in terest ing and important to

note that th is in tegrat ion, and re la ted in format ion feedback respect ive ly, is s t i l l most ly

l imi ted to “system-externa l ” aspects such as, for example, env i ronmenta l or economic

cr i ter ia , whereas the “system- int r ins ic” mater ia l character is t ics and phys ica l behav ior

are hard ly ever cons idered or ut i l i zed (Menges 2008).

The la rger research pro jec t th is paper is based on a ims fo r deve lop ing a computa t iona l

des ign approach tha t syn thes izes per fo rmance-or ien ted fo rm genera t ion and

phys ica l p rocesses o f mater ia l i za t ion. Here, the des ign space is de f ined and

const ra ined by mater ia l behav io r, fabr ica t ion and product ion. Th is unders tand ing o f

des ign computa t ion as a ca l ib ra t ion between the v i r tua l p rocesses o f genera t ing

fo rm and the phys ica l becoming o f mater ia l sys tems, shou ld not be conce ived as

l im i t ing the des igner, but ra ther as enab l ing the exp lo ra t ion o f unknown po in ts in

the search space def ined by the mater ia l i t se l f (Menges 2010) . In o ther words, th is

research s t r i ves fo r cont r ibu t ing to the fu r ther deve lopment o f computa t iona l des ign

by invest iga t ing [ i ] how mater ia l behav io r can be in tegra ted in genera t i ve des ign

computa t ion, [ i i ] how th is in tegra t ion requ i res not mere ly a CAD-CAM cha in but ra ther

an in fo rmat ion feedback between the degrees o f f reedom and const ra in ts o f robot ic

manufactu r ing and mater ia l a f fo rdances, and [ i i i ] how such processes can prov ide fo r

the deve lopment o f nove l , per fo rmat i ve a rch i tec tu ra l mate r ia l sys tems.

2 Integrat ing Computat ional Form Generat ion and Physical Behaviour

Based on the a fo rement ioned la rger research agenda, th is paper w i l l concent ra te

on par t icu la r aspects o f mater ia l behav io r, the actua l mate r ia l make-up dete rmin ing

th is behav io r, and the fabr ica t ion processes tha t a l low fo r exp lo i t ing th is mater ia l

behav io r. The focus w i l l be on the e las t ic bend ing behav io r o f robot ica l l y fabr ica ted

wood e lements . The reasons fo r th is a re th ree fo ld and exp la ined in the fo l low ing

paragraphs:

1. E las t ic bend ing as both a fo rm- f ind ing and const ruc t ion techn ique is s t i l l

re la t i ve ly uncommon. Desp i te the cons iderab le per fo rmat i ve capac i ty o f e las t ica l l y

bent s t ruc tu res, th is may be due to the fac t tha t a rch i tec ts and eng ineers a l i ke

s t i l l l ack too ls fo r des ign ing w i th geomet r ica l l y ins tab le e lements . A lso, s t ruc tu ra l

eng ineers a re genera l l y t ra ined to unders tand la rger de fo rmat ions as prob lemat ic

and potent ia l l y damag ing. Because o f the cons iderab le techn ica l and in te l lec tua l

d i f f i cu l t ies posed by a synchronous cons idera t ion o f fo rce, fo rm and per fo rmance,

there a re on ly ve r y few cases o f e las t ica l l y -bent a rch i tec tu res, as fo r example the

Madan peop le ’s ve rnacu la r s t ruc tu res o f bent reed bund les in I raq (F igure 1a) , o r

the wooden she l l s t ruc tu res a t Hooke Park in Eng land (F igures 1b/c) , wh ich were

co l labora t i ve ly des igned by F re i Ot to , ABK and Buro Happo ld , and const ruc ted f rom

e las t ica l l y bent g reen round-wood po les prov ided by loca l fo res t th inn ing.

Page 3: Integrative Design Computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf · 2011-09-13 · Integrative Design Computation integrating material behaviour and robotic

74

integration through computationacadia 2011 _proceedings

Figure 2a. White Oak (Copley 2011)

Figure 2b. Wood Cell (Dinwoodie 2000)

Figure 2c. Anisotropic E-Modulus (Dinwoodie 2000)

Fig. 2a

Fig. 2b

Fig. 2c

2. Wood is a complex mater ia l . In contrast to other more homogenous and isotropic

bui ld ing mater ia ls speci f ical ly produced for the demands of the bui ld ing sector, wood is a

natura l ly grown bio logical t issue character ized by a di f ferent iated f ibrous structure. Whi le

i t is commonly accepted by archi tects, engineers and craf tsmen al ike that the speci f ic

mater ia l make-up of wood needs to be considered in the design process, archi tectura l

design tools are not capable of integrat ing anisotropic mater ia l character ist ics and

behavior. Thus the complex anatomy of wood and i ts re lated behavior is of part icular

interest for an informat ion-based, computat ional design process.

3. Robot ic manufactur ing opens up the possibi l i ty of combining computer control led

6-axis k inemat ics with re lat ively common tools for wood working as, for example, rotary

saw blades and mi l l ing cutters. In the aforement ioned integrat ive computat ional design

process the large design space af forded by the var ious degrees of f reedom of the

industr ia l robot and i ts ef fector a l lows for speci f ical ly di f ferent iat ing bui ld ing elements in

the prefabr icat ion process, and thus extending the complex micro scale structure of the

mater ia l to the macro scale of the mater ia l construct ion. In th is way, the possible macro-

scalar di f ferent iat ion dur ing the manufactur ing process and consequent ia l manipulat ion

of the mater ia l e lement’s behavior needs to be integrated in the computat ional design

process. In addit ion, as the programming of a 6-axis robot is re lat ively complex and

t ime-consuming, the data output in robot control language needs to be an integral part

of the informat ion model for the fabr icat ion of large numbers of di f ferent iated elements.

Consider ing these three points the paper wi l l f i rst expla in the part icular anatomy of

wood, and how the microscopic f ibrous structure of wood governs the macroscopic

bending behavior of wood pieces. Thereafter, the presentat ion of three research projects

demonstrates di f ferent approaches to the computat ional design integrat ion: The f i rst

project explores the computat ional integrat ion of mater ia l and fabr icat ion parameters for

plast ical ly formed hardwood slats. For the construct ion of an intr icate, l ightweight lat t ice

structure the process var iables of steaming, bending and twist ing s lender white oak

members were embedded in a computat ional design tool. The second project invest igates

the possibi l i ty of programming the elast ic bending behavior of pre-steamed hardwood

slats by local ly di f ferent iated kerfs. In order to construct a fu l l scale, hyperboloid

prototype structure the mater ia l propert ies of whi te oak, the process var iables of wood

steaming and the fabr icat ion parameters of a 6-axis industr ia l robot equipped with a

ci rcular saw blade were integrated in a custom programmed design tool. This enabled

both the computat ional form-f inding of the system and the automat ic generat ion of the

re lated machine, which was di rect ly der ived as the required robot control language. The

th i rd project researches a computat ional design process based on the elast ic bending

behavior of b i rch plywood lamel las. These lamel las are robot ical ly manufactured as

in i t ia l ly p lanar e lements, and subsequent ly connected so that the force that is local ly

stored in each bent region of the str ip, and mainta ined by the corresponding tensioned

region of the neighbor ing str ip. The resul tant novel bending-act ive structure uses only

extremely th in plywood lamel las, which sett le into an equi l ibr ium state that unfolds a

unique archi tectura l space whi le at the same t ime being very ef f ic ient wi th the employed

mater ia l resources.

3 Integrat ing the Fibrous Structure, Di fferent iated Anatomy and

Anisotropic Behavior of Wood

As a natura l ly grown t issue, wood is s ign i f icant ly d i f fe rent to most other mater ia ls used

in the bu i ld ing indust r y. Whi le wood is one of the o ldest const ruct ion mater ia ls , i ts

organ ic nature is aga in understood as a s ign i f icant advantage, par t icu lar ly in the l ight

of the future env i ronmenta l cha l lenges of the bu i ld ing sector. As t rees main ly ut i l i ze

photosynthes is, that is so lar energy, for growth, and as they t ransform carbon d iox ide

in to oxygen dur ing th is processes, wood has an ext remely low leve l o f embodied

energy and a pos i t ive CO2 ba lance, even af ter undergoing today’s h igh ly indust r ia l

wood process ing. Thus, i f grown in susta inable s i lv icu l ture, wood is one of the very few

h igh ly energy ef f ic ient , natura l ly renewable, fu l ly recyc lab le bu i ld ing mater ia ls (Herzog

2003). As a consequence, a computat iona l approach to des ign ing wood const ruct ions

is of par t icu lar re levance. Even more so as the complex st ructure and re la ted mater ia l

behav ior of wood, which a lso or ig inates f rom i ts organ ic nature, lends i tse l f to an

Page 4: Integrative Design Computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf · 2011-09-13 · Integrative Design Computation integrating material behaviour and robotic

75

Figure 3a. Steaming white oak slats

Figure 3b. Steamed slat forming figure

Fig. 3a

Fig. 3b

in format ion-based des ign process. Hence, any genuine wood-speci f ic computat iona l

des ign process needs to begin wi th an understanding of wood’s anatomy, as i t

accounts for most of i ts proper t ies and character is t ics (Hensel and Menges 2009).

Wood grows as the f ibrous, vascular t issue of t rees funct ion ing as both the load-

bear ing st ructure and metabol ic in f rast ructure for the l i v ing p lant organ isms. The

anatomy of wood has evo lved in response to these b io log ica l requi rements and thus

d i f fers cons iderab ly f rom other const ruct ion mater ia ls that are indust r ia l l y produced

and des igned to sat is fy the speci f ic demands of the bu i ld ing indust r y. These mater ia ls ,

for example stee l and g lass, are typ ica l ly homogenous, i .e. un i form in composi t ion,

and isot rop ic, hav ing equal or very s imi la r proper t ies in a l l d i rect ions. In cont rast ,

wood is not on ly subject to b io log ica l var iab i l i ty and natura l i r regu lar i t ies, wood’s

mater ia l make-up i tse l f is grown as a heterogeneous, an isot rop ic and h ierarch ica l

s t ructure. The reproduct ive t issues of t rees grow both on the t ips of tw igs (ap ica l

mer is tems) and as th ickness growth of the stem ( la tera l mer is tems) adding year ly

wood to the prev ious ly grown. The th ickness growth pr imar i ly takes p lace in a very th in

layer of ce l ls , the cambium between the inner s tem and the bark. Cambia l ce l ls d iv ide

to form another cambia l ce l l and a new ce l l that , depending on i ts pos i t ion, e i ther

matures as a bark or wood ce l l (Hoadley 2000). The ce l lu la r s t ructure of sof twood and

hardwood is re la t ive ly d i f fe rent . Whereas more than 90 percent of sof twood’s t issue

is composed of t racheids, which are long f iber- l ike ce l ls ar ranged para l le l to the stem

ax is, hardwood cons is ts of numerous ce l l types inc lud ing t racheids, vesse ls, rays and

f ibre ce l ls (Wagenführ 1999). The b i rch and whi te oak wood (F igure 2a) used in the

three research pro jects d iscussed in th is paper be long to the c lose ly re la ted hardwood

fami l ies of Betu laceae and Fagaceae.

Af ter the wood ce l ls are fu l ly grown to the i r f ina l s ize and shape, the inner sur face of

the f rag i le pr imary ce l l wa l l is re in forced by a s ign i f icant ly th icker secondary ce l l wa l l

cons is t ing of three layers (F igure 2b) . The ce l l wa l ls are const ructed f rom long cha in

ce l lu lose molecules main ly or iented para l le l to the ce l l ’s long ax is. The l inear po lymer

cha in of ce l lu lose is h igh ly ordered wi th densely packed molecules, resu l t ing in f ibrous-

l ike st rands ca l led microf ibr i ls . Th is ce l lu los ic st ructure is then re in forced by a matr ix o f

l ign in. The resu l t ing d i rect iona l composi te system accounts for many character is t ics of

wood (Barnet t and Jeron imid is 2003). The h igh ly d i rect iona l , para l le l a l ignment of the

ce l lu lose cha ins in the microf ibr i ls resu l ts in s t rongly an isot rop ic mechanica l proper t ies

(F igure 2c) . For example, wood has h igh tens i le and compress ive st rength para l le l

to the gra in d i rect ion, whereas i t is very low in compress ion or tens ion perpendicu lar

to the f ibre. S imi la r ly, wood’s modulus of e last ic i ty d i f fe rs s ign i f icant ly depending on

gra in or ientat ion: the modulus of e last ic i ty para l le l to the main f ibre d i rect ion, between

9000 to 16000 N/mm² depending on wood species, is genera l ly approx imate ly f i f teen

t imes h igher than perpendicu lar to the f ibres, which is between 600 to 1000 N/mm²

(D inwoodie 2000). Th is can be understood and inst rumenta l i zed as an in terest ing

proper ty of var iab le st rength and st i f fness in re la t ion to gra in or ientat ion (Wagenführ

2008).

S imi la r to synthet ic composi tes, e.g. , g lass f ibre re in forced p last ic, wood is

character ized by re la t ive ly h igh st ra in at fa i lu re, which means h igh load-bear ing

capaci ty wi th re la t ive ly low st i f fness. These mater ia l proper t ies and re la ted behav ior

are especia l ly wel l su i ted for const ruct ion techniques that employ the e last ic bending

of wood in order to form complex, l ightweight s t ructures f rom in i t ia l l y s imple, p lanar

bu i ld ing e lements, as the fo l lowing three pro jects wi l l demonst rate.

4 Performative Wood Construct ion

4.1 RESEARCH PROJECT 01: LATTICE SYSTEM CONSTRUCTED FROM STEAMED WHITE OAK

Bend ing so l i d wood i s a t r ad i t i ona l woodwo rk i ng t echn ique . As compa red

t o add i t i v e o r sub t r ac t i ve f ab r i ca t i on t echn iques , t h i s f o rm ing p rocess has

cons ide rab l e advan tages f o r p roduc i ng cu r ved wooden pa r t s : bend i ng wood i s

ma te r i a l l y ve r y e f f i c i en t and s t r uc tu r a l l y advan tageous , as i t r eo r i en t s t he g r a i n

d i r ec t i on t o f o l l ow t he pa r t ’s cu r va tu re , a vo id i ng excess i ve f i be r r un -ou t on t he

edges and c ross -g ra i n weaknesses .

computation, formation and materiality

Page 5: Integrative Design Computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf · 2011-09-13 · Integrative Design Computation integrating material behaviour and robotic

76

integration through computationacadia 2011 _proceedings

Figure 4a. Computationally derived prototype

set-up

Figure 4b/c. Physical prototype structure based on

the bending behavior of steamed white oak slats

Fig. 4a

Fig. 4b Fig. 4c

The l imits of bending solid wood are defined by beam mechanics. In a bent beam, the

outer f ibers on the concave side are in compression and on the convex side in tension,

usually fai l ing at approximately one percent elongation. As the material properties of wood

are dependent on temperature and moisture content, a tradit ional method for extending the

plasticity of a number of hardwood species e.g., ash, beech, birch, elm and white oak, is

steaming. When the wood reaches moisture content close to the fiber saturation point, the

steam softens the wood fibers and allows them to distort in relation to each other (Keyser

1985). This increases the elongation abil ity in tension to approximately two percent and

the compressibil i ty to a signif icant 30 percent or more. Thus steaming allows extending the

range of possible curvatures considerably and, if the steaming time is l imited to the required

one to two minutes per mil l imeter thickness at approximately 100 to 110° C, the strength

losses do not exceed 20 percent (Hoadley 2000).

Tradit ionally, steam-bending processes require a mould that defines the curvature of the

part to be formed. This project, developed by Jeffrey Niemasz, Jon Sargent and Laura

Viklund (Performative Wood Studio, Prof. Achim Menges, Harvard GSD 2009), explored

ways of developing a computational design process that al lows the free forming of steamed

white-oak slats without the need for a 100 percent contact mould. In order to achieve this,

a large number of physical tests were conducted, investigating the process parameters of

steaming e.g. temperature, t ime (Figure 3a), the specif ication parameters of the slats such

as species, f iber direction, init ial moisture content, and thickness, in relation to the forming

figure (Figure 3b) that the slats f ind when the posit ion of one endpoint is translated in space

in relation to the other. The resultant data was used to set up a computational form-finding

tool for a multi layered latt ice system constructed from bent and twisted white oak slats.

The deve loped in tegra t i ve computa t iona l des ign too l was tes ted by const ruc t ing a

fu l l sca le p ro to type f rom more than 1000 geomet r ica l l y un ique par ts . The l igh twe ight

s t ruc tu re is computa t iona l l y der i ved as a doub le layered la t t ice prov id ing both a

Page 6: Integrative Design Computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf · 2011-09-13 · Integrative Design Computation integrating material behaviour and robotic

77

Figure 5a. Kerfed oak

Figure 5b. Kerf forming

Figure 5c. Kerf variables

Figure 6a. Gradual kerf orientation/depth changes

(top) avoiding kinks (bottom)

Figure 6b. Bending behavior (top) of robot-sawn

gradual kerf depth (bottom)

Fig. 5a Fig. 5b

Fig. 5c

spat ia l d iv is ion and a seat ing sur face, in wh ich the th ickness o f s la ts , the number

o f layers and the depth o f the sys tem responds to the load-bear ing requ i rements

(F igure 4a) . Ut i l i z ing a custom-bu i l t s team chamber and a s imp le CNC-cut fo rmwork

tha t gu ides the f ree fo rming o f the s teamed p ieces by spat ia l re fe rence po in ts , the

pro to type was const ruc ted f rom wh i te oak s la ts w i thout any add i t iona l mechan ica l

fas teners o r f i x ings (F igures 4b/c) .

4.2 RESEARCH PROJECT 02: HYPERBOLOID STRUCTURE CONSTRUCTED FROM PRE-STRESSED

ROBOTICALLY KERFED WHITE OAK ELEMENTS

The nex t r esea rch p ro j ec t , deve l oped by B rad C rane , And rew McGee , Ma rsha l l

P r ado and Yang Zhao (Pe r f o rma t i ve Wood S tud io , P ro f . Ach im Menges , Ha r va rd

GSD 2010 ) , i n ves t i ga ted ways o f e x t end i ng t he poss ib i l i t y o f s t eamed , f r ee f o rmed

wooden s l a t s t h rough s t r a t eg i c accumu la t i ve l oca l weaken i ng and d i s r up t i on o f

f i be r con t i nu i t y by ke r f i ng ( F igu re 5a ) . The p ro j ec t a imed to deve lop an i n t eg ra ted

compu ta t i ona l des i gn t oo l and robo t i c manu fac tu r i ng p rocess t ha t a l l ows

p rog ramming t he bend ing and tw i s t i ng behav i o r ( F igu re 5b ) o f t ens i oned wood

e l emen ts t h rough spec i f i c ke r f i ng ( F igu re 5c ) .

Because o f wood ’s an i so t r op i c cha rac te r i s t i c s , ma te r i a l pe rpend icu l a r t o t he ma in

g r a i n d i r ec t i on can be r emoved w i t hou t ove r l y comprom is i ng t he ove ra l l s t r uc tu r a l

capac i t y. I n boa t cons t r uc t i on , f u r n i t u r e mak i ng and o the r f i e l ds , r egu l a r ke r f i ng

i s a we l l known t echn ique f o r f ab r i ca t i ng wooden pa r t s ben t i n one d i r ec t i on .

Th i s p ro j ec t e xp lo red how the compu te r con t r o l l ed va r i a t i on o f ke r f dep th , l eng th ,

f r equency and o r i en t a t i on a l l owed f o r ach i ev i ng mo re e l abo ra te bend ing and

wa rp i ng f i gu res .

Unde rs tood as a s ys tem o f cumu la t i ve ke r f s , t he mac ro -sca l e man ipu l a t i on o f t he

wooden s l a t s p r i o r t o s t eam ing mod i f i e s t he i r bend i ng behav i o r. Cons tan t ke r f

dep th r esu l t s i n s t r ess concen t r a t i on a t t he end o f ke r f ed l eng th , l ead i ng t o t he

i so l a t ed ac t i v a t i on o f t hese r eg i ons and consequen t l y p roduces k i nks a t t hese

po i n t s ( F igu re 6a ) . Howeve r, va r y i ng ke r f dep th g r adua l l y i n r e l a t i on t o t he s t r ess

d i s t r i bu t i on a l l ows f o r ca l i b r a t i ng t he bend ing s t i f f ness w i t h ma te r i a l r emova l . Fo r

e xamp le , i f t he dep th va r i a t i on o f pa ra l l e l ke r f s f o l l ows a s i ne cu r ve , t he r esu l t an t

f i gu re d i sp l a ys g r adua l cu r va tu re change avo id i ng s t r ess concen t r a t i ons o r k i n ks

( F igu re 6b ) . Robo t i c saw ing p rov i des t he r equ i r ed va r i ab i l i t y and p rec i s i on t o

i ns t r umen ta l i z e ke r f i ng i n t h i s way.

A cus tom-des igned ro t a r y saw too l f o r a 6 -ax i s r obo t ( F igu re 7a ) was cons t r uc ted

and enab l ed t he t es t i ng o f r e l a t ed p rocess pa rame te r s , e .g . , saw b l ade r evo l u t i on ,

f eed r a t e , c l imb cu t t i ng and conven t i ona l cu t t i ng , e t c . , i n r e l a t i on t o geome t r i c ke r f

pa rame te r s i nc l ud i ng ke r f dep th , l eng th , f r equency and o r i en t a t i on ( F igu re 7b ) .

The behav i o r a l cha rac te r i s t i c s o f t he r esu l t an t t es t p i eces ( F igu re 7c ) t oge the r

w i t h t he f ab r i ca t i on pa rame te r s we re i n t eg ra ted i n a compu ta t i ona l des i gn t oo l and

t es ted t h rough t he cons t r uc t i on o f a l a r ge r sca l e p ro to t ype .

As i n i t i a l t es t s had shown t ha t spec i f i c ke r f pa t t e r ns a l l ow f o r ach i ev i ng a

s ys tem-geome t r y w i t h nega t i ve Gauss i an geome t r y, once an assemb l y o f mu l t i p l e

ke r f ed e l emen ts i s p re -s t r essed , t he p ro to t ype was deve loped as an i r r egu l a r

h ype rbo lo i d g l oba l s ys tem cons i s t i ng o f mo re t han 140 e l emen ts w i t h un ique l oca l

ke r f pa t t e r ns ( F igu re 8a ) . Cond i t i on i ng t he phys i ca l f o rm ing behav i o r o f each

e l emen t , t he compu ta t i ona l des i gn t oo l gene ra tes t he i nd i v i dua l ke r f pa t t e r ns ,

p rov i des t he r e l e van t geome t r i c da t a and ou tpu t s t h i s d i r ec t l y i n r obo t con t r o l

code . Th i s enab l es t he d i r ec t f ab r i ca t i on o f t he i nd i v i dua l ke r f ed p i eces t ha t we re

subsequen t l y s t eamed , p re -s t r essed , assemb led i n t o componen ts and f i na l l y

assemb led as a 5 me te r t a l l h ype rbo lo i d p ro to t ype ( F igu res 8b /c ) .

4.3 RESEARCH PROJECT 03: BENDING-ACTIVE PAVILION CONSTRUCTED FROM ROBOTICALLY

FABRICATED AND ELASTICALLY BENT BIRCH PLYWOOD LAMELLAS

This research project conducted by the the Inst i tute for Computat ional Design (Prof. Achim

Menges) and the Inst i tute of Bui lding Structures and Structural Design (Prof. Jan Knippers)

Fig. 6a

Fig. 6b

computation, formation and materiality

Page 7: Integrative Design Computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf · 2011-09-13 · Integrative Design Computation integrating material behaviour and robotic

78

integration through computationacadia 2011 _proceedings

Figure 7a. Robot-saw tool

Figure 7b. Robotic kerfing

Figure 7c. Variable kerfs

Figure 8a. Prototype

Figure 8b/c. Prototype structure and kerfing detail

Fig. 7a

Fig. 7b Fig. 7c

a imed a t fu r ther deve lop ing mater ia l -based computa t iona l des ign, eng i nee r i ng and

robo t i c manu fac tu r i ng o f e l as t i ca l l y ben t wooden s t r uc tu res . The r esu l t i s a nove l

p re -s t r essed s t r uc tu r a l s ys tem - he re r e f e r r ed t o as a bend ing -ac t i ve s t r uc tu re

( L i enha rd e t a l . 2010 ) - an i n t r i ca te ne two rk o f j o i n t po i n t s and r e l a t ed f o rce

vec to r s t ha t a r e spa t i a l l y med i a t ed by t he e l as t i c ma te r i a l behav i ou r o f t h i n b i r ch

p l ywood l ame l l a s , wh i ch was deve loped and subsequen t l y t es ted as a f u l l sca l e

r esea rch pav i l i on .

The bas i c p r i nc ip l e o f t h i s s ys tem i s t ha t t he i n i t i a l l y p l ana r p l ywood s t r i ps a re

manu fac tu red w i t h a 6 -ax i s i ndus t r i a l r obo t and subsequen t l y connec ted so t ha t

e l as t i ca l l y ben t and t ens i oned r eg i ons a l t e r na te a l ong t he i r l eng th . The f o rce

t ha t i s l oca l l y s t o red i n each ben t r eg i on o f t he s t r i p , and ma in t a i ned by t he

co r r espond ing t ens i oned r eg i on o f t he ne i ghbo r i ng s t r i p , g rea t l y i nc reases t he

s t r uc tu r a l capac i t y o f t he sys tem ( F igu re 9a ) . I n o rde r t o p reven t l oca l s t r ess

concen t r a t i ons as we l l a s t he ad j acency o f weak spo ts w i t h i n t he ove ra l l s ys tem,

t he l oca t i ons o f t he j o i n t s be tween connec ted s t r i ps need to osc i l l a t e a l ong t he

s t r uc tu re , r esu l t i ng i n a d i s t i nc t a r t i cu l a t i on o f t he enve l ope . Th i s compu ta t i ona l l y

gene ra ted , i r r egu l a r d i s t r i bu t i on o f l oca l j o i n t po i n t s ( F igu re 9b ) g rea t l y enhances

t he s t r uc tu r a l capac i t y o f t he g l oba l s ys tem, bu t a l so r equ i r es each pa r t t o be

geome t r i ca l l y un ique .

The des i gn o f t he p ro to t ype pav i l i on began w i t h t he deve lopmen t o f a compu ta t i ona l

des i gn t oo l . I n t h i s t oo l a l l r e l e van t ma te r i a l behav i o r a l cha rac te r i s t i c s a re

i n t eg ra ted as pa rame t r i c dependenc i es based on a l a rge numbe r o f phys i ca l

and compu ta t i ona l t es t s . These t es t s f ocused on measu r i ng t he de f l ec t i ons o f

e l as t i ca l l y ben t p l ywood s t r i ps i n r e l a t i on t o va r i ous spec i f i ca t i on pa rame te r s as

we l l a s t he ca l i b r a t i on and co r robo ra t i on o f t he r esu l t i ng da ta w i t h f i n i t e e l emen t

me thods ( FEM) ( F igu re 10a ) . Based on 6400 l i nes o f code , t he deve loped

i n t eg ra t i ve compu ta t i ona l t oo l gene ra tes poss ib l e s ys tem mo rpho log i es t oge the r

w i t h a l l r e l e van t geome t r i c i n f o rma t i on and d i r ec t l y ou tpu t s t he da ta r equ i r ed f o r

bo th subsequen t FEM s imu l a t i ons and t he manu fac tu r i ng w i t h a 6 -ax i s i ndus t r i a l

r obo t ( F igu re 10b ) .

Embedded i n t he compu ta t i ona l des i gn p rocess , t he FEM s imu l a t i ons we re used

to ca l cu l a t e t he ac tua l ma te r i a l behav i ou r unde r t he g i ven geome t r i c and phys i ca l

cond i t i ons wh i l e cons ide r i ng a l l ac t i ng f o rces and ma te r i a l l im i t a t i ons . I n o rde r t o

s imu l a t e t he i n t r i ca te g l oba l equ i l i b r i um o f l oca l l y s t o red ene rgy t ha t r esu l t s f r om

the bend ing o f each s t r i p , t he mode l needs t o beg i n w i t h t he p l ana r d i s t r i bu t i on

o f t he 80 s t r i ps ( F igu re 11a ) , f o l l owed by s imu l a t i ng t he e l as t i c bend ing and

subsequen t coup l i ng o f t he s t r i ps t o f o rm a comb ined se l f - s t ab i l i z i ng s t r uc tu re

( F igu re 11b ) . The FEM mode l a l l owed f o r ve r i f y i ng t he geome t r i ca l shape w i t h i n

p rede f i ned s t r ess l e ve l s and ma te r i a l capac i t y u t i l i z a t i on , as we l l a s ana l y z i ng t he

de fo rma t i ons and s t r ess d i s t r i bu t i ons unde r ex te rna l w i nd l oads .

I n add i t i on t o t he ma te r i a l behav i ou r, t he manu fac tu r i ng and assemb l y l og i cs we re

i n t eg ra ted i n t he compu ta t i ona l p rocess . Based on t he mach i ne cons t r a i n t s o f

t he 6 -ax i s f ab r i ca t i on r obo t t o be used ( F igu re 12a ) , t he t h r ee c r i t i ca l de t a i l s o f

t he sys tem we re deve loped : [ i ] t he shea r- r es i s t an t j o i n t f o r connec t i ng ad j acen t

s t r i ps , [ i i ] t he t ens i on puzz l e j o i n t t o connec t s t r i p segmen ts o f l im i t ed s tock

s i ze ( F igu re 12b ) , and [ i i i ] t he j o i n t be tween e l as t i c s t r i p and t he s t r uc tu r a l base

( F igu re 12c ) . The d i r ec t gene ra t i on o f a l l manu fac tu r i ng da ta a l l owed t he r ap id

f ab r i ca t i on o f 500 geome t r i ca l l y un ique pa r t s .

The assembly p rocess is s t ra igh t fo rward and qu ick to execute , w i th no need fo r

ex tens ive sca f fo ld ing or add i t iona l equ ipment , as the p lanar s t r ips s imp ly need to

be connected and then automat ica l l y f ind the i r spec i f ic shape (F igure 13a) . In o ther

words, the mater ia l i t se l f u l t imate ly computes the shape o f the pav i l ion (F igure 13b) .

The spa t i a l a r t i cu l a t i on and s t r uc tu r a l s ys tem i s based on a ha l f - t o r us shape .

De f i n i ng t he u rban edge o f t he s i t e , i t t ouches t he g round topog raphy t ha t

p rov i des sea t i ng oppo r t un i t i e s on t he s t r ee t f ac i ng co rne r. I n con t r as t t o t h i s , t he

t o rus s i de t ha t f aces t he pub l i c squa re i s l i f t ed f r om the g round to f o rm a f r ee -

Fig. 8a Fig. 8b

Fig. 8c

Page 8: Integrative Design Computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf · 2011-09-13 · Integrative Design Computation integrating material behaviour and robotic

79

spann i ng open i ng . I n s i de , t he t o ro i da l space can neve r be pe rce i ved i n i t s en t i r e t y

( F igu re 14a ) , l ead i ng t o a su rp r i s i ng spa t i a l dep th t ha t i s f u r t he r enhanced by

t he sequence o f d i r ec t and i nd i r ec t i l l um ina t i on r esu l t i ng f r om the convex and

concave undu l a t i ons o f t he enve l ope . I n t he f i na l s ys tem, t he comb ina t i on o f t he

p re -s t r ess r esu l t i ng f r om the e l as t i c bend ing du r i ng t he assemb l y p rocess and

t he mo rpho log i ca l d i f f e r en t i a t i on o f t he j o i n t l oca t i ons enab l es a ve r y l i gh twe igh t

s ys tem. The en t i r e pav i l i on , w i t h a d i ame te r o f mo re t han twe l ve me te r s , can be

cons t r uc ted us i ng ex t r eme l y t h i n b i r ch p l ywood l ame l l a s w i t h a t h i ckness o f on l y

6 .5mm. Th i s ex t r eme l y t h i n and ma te r i a l l y e f f i c i en t s k i n se r ves s imu l t aneous l y

as t he l oad -bea r i ng s t r uc tu re as we l l a s t he l i gh t modu l a t i ng and r a i n p ro tec t i ng

enve l ope ( F igu re 14b ) f o r t he sem i - i n t e r i o r e x t ens i on o f t he pub l i c squa re .

Beyond the invest iga t ion o f the re la ted a rch i tec tu ra l qua l i t i es , the const ruc t ion o f the

research pro jec t a l lowed fo r ve r i f y ing the presented computa t iona l des ign approach

by compar ing the computa t iona l des ign mode l , the re la ted FEM mode l and the actua l

geomet r y o f the const ruc ted pav i l ion . In co l labora t ion w i th geodes ic eng ineers , the

pav i l ion , as const ruc ted on s i te , was repeated ly d ig i t i zed us ing fu l l sca le scann ing

and geodes ic measurement techn iques (F igure 15a) resu l t ing in po in t -c loud

dataset . These exact measurements and re la ted mode ls a l lowed fo r compar ing the

computa t iona l des ign mode l , the FEM s imu la t ion and the actua l l y bu i l t s t ruc tu re ,

showing on ly m inor dev ia t ions between the th ree datasets (F igure 15b) .

I n add i t i on , t he scann i ng enab l ed t he obse r va t i on o f t he pav i l i on ’s s t r uc tu r a l

pe r f o rmance ove r a l onge r t ime pe r i od by documen t i ng t he geome t r y changes

r esu l t i ng f r om the r e l a xa t i on and c reep i ng o f t he p l ywood s t r i ps . I n t he f u t u re t h i s

i n f o rma t i on can be used to ca l i b r a t e t he FEM s imu l a t i on and he lps t o p red i c t t he

l ong t e rm behav i o r o f e l as t i ca l l y ben t wood cons t r uc t i ons .

5 Conclusion

The t h ree r esea rch p ro j ec t s demons t r a t e how a syn thes i s o f ma te r i a l , f o rm and

pe r f o rmance i n i n t eg ra t i ve compu ta t i ona l des i gn p rocesses a l l ows f o r de r i v i ng

comp lex s t r uc tu res f r om uncomp l i ca ted ma te r i a l s ys tems , wh i ch a re bo th

econom ica l t o bu i l d and ma te r i a l l y e f f i c i en t , wh i l e a t t he same t ime p rov i d i ng

un ique a rch i t ec tu r a l oppo r t un i t i e s . Th i s becomes pa r t i cu l a r l y c l ea r i n t he l a s t

p ro j ec t : Compa r i ng t he r esu l t s o f t he gene ra t i ve compu ta t i ona l des i gn p rocess

w i t h FEM s imu l a t i ons and t he exac t measu remen t o f t he geome t r y t ha t t he ma te r i a l

“ compu ted ” on s i t e i nd i ca tes t ha t t he sugges ted i n t eg ra t i on o f des i gn compu ta t i on

and ma te r i a l i z a t i on i s no l onge r an i dea l i z ed goa l bu t a f eas ib l e p ropos i t i on .

Acknowledgements

The au tho r g r a t e f u l l y acknow ledges t he r esea rch wo rk o f a l l pa r t i c i pan t s i n t he

Pe r f o rma t i ve Wood s tud ios a t Ha r va rd GSD, and i n pa r t i cu l a r t he s i gn i f i can t

con t r i bu t i ons o f Je f f r e y N i emasz , Jon Sa rgen t , Lau ra V i k l und and B rad C rane ,

And rew McGee , Ma rsha l l P r ado , and Yang Zhao . I n add i t i on , t he au tho r t hanks

h i s co l l eague Jan Kn ippe rs and a l l ICD and I TKE r esea rche rs and s tuden ts who

pa r t i c i pa ted i n t he Resea rch Pav i l i on p ro j ec t a t S tu t t ga r t Un i ve r s i t y, i n pa r t i cu l a r

Mo r i t z F l e i schmann , S imon Sch l e i che r, Ch r i s t ophe r Robe l l e r, Ju l i an L i enha rd ,

D i ana D ’Souza , Ka ro l a D i e r i chs as we l l a s And reas E i senha rd t , Manue l Vo l l r a t h ,

K r i s t i ne Wäch te r, Thomas I r owe t z , O l i ve r Dav id K r i eg , Ádm i r Mahmu tov i c , Pe te r

Meschendö r f e r, Leopo ld Möh le r, M ichae l Pe l ze r and Kon rad Ze rbe .

Fig. 9b

Fig. 9a

Figure 9a. Two jointed plywood strips with

alternating tensioned/bent segments

Figure 9b. Computationally derived distribution of

joint points

Figure 10a. Phyical / FEM bending tests

Figure 10b. Computational model

Figure 11a. FEM planar strip layout

Figure 11b. Equilibrium state simulation

Fig. 10a

Fig. 10b

Fig. 11a Fig. 11b

computation, formation and materiality

Page 9: Integrative Design Computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf · 2011-09-13 · Integrative Design Computation integrating material behaviour and robotic

80

integration through computationacadia 2011 _proceedings

Figure 12a. Robotic strip fabrication

Figure 12b. Tension joint detail

Figure 12c. Base joint detail

Figure 13a. Assembly of pavilion

Figure 13b. Bent plywood strip pavilion

Fig. 12a

Fig. 12b

Fig. 12c

Fig. 13a

Fig. 13b

Page 10: Integrative Design Computationpapers.cumincad.org/data/works/att/acadia11_72.content.pdf · 2011-09-13 · Integrative Design Computation integrating material behaviour and robotic

81

Figure 14a. Interior view of pavilion

Figure 14b. Exterior view of pavilion

Figure 15a. Scanning of pavilion

Figure 15b. Dataset comparison

Fig. 14a

Fig. 14b

Fig. 15a

Fig. 15b

References

Barne t t J . R . and Je ron im id i s G. Eds . 2003. Wood qua l i t y and i t s b io log ica l bas is .

Ox fo rd : B lackwe l l CRC Press .

D inwood ie J . M. 2000. T imber : I t s na tu re and behav iou r. London: E&FN Spon.

Hense l , M. and A . Menges . 2009. Ho l z Fo rm F indung. A rchP lus , No . 193, 106-109.

Herzog T. 2003. Ho l zbau A t l as . Base l : B i r khäuse r.

Hoad ley B . 2000. Unders tand ing wood. 2nd Ed i t i on . Newtown: Taun ton P ress .

Howard , R . 1998. Comput ing i n cons t ruc t ion : P ionee rs and the fu tu re . Ox fo rd /

Woburn : Bu t te rwor th -He inemann.

Keyse r, W. 1985. S teambend ing . I n Bend ing Wood, 16 – 22 . Newtown: Taun ton

P ress .

L ienha rd , J . , S . Sch le iche r, J . Kn ippe rs , S . Popp inga , and T. Speck . 2010. Fo rm-

f i nd ing o f na tu re i nsp i res k inemat ics fo r p l i ab le s t ruc tu res . I n Zhang , Q. e t a l .

( eds ) P roceed ings o f the I n te rna t iona l Sympos ium o f the I n te rna t iona l Assoc ia t i on o f

She l l and Spa t i a l S t ruc tu res ( IASS) , Spa t i a l S t ruc tu res Tempora r y and Pe rmanen t ,

Shangha i , Ch ina , 501 - 507.

Mark , E . , M. Gross , G. Go ldschmid t . 2008. A pe rspec t i ve on Compute r A ided

Des ign a f te r fou r decades . A rch i tec tu re i n Comput ro , P roceed ing o f the 26 th

eCAADe Con fe rence , An twerpen (Be lg ium) 17-20 Sep tember 2008, 169 - 176.

Menges , A . 2010. I n teg ra l computa t i ona l des ign : Syn thes i z i ng computa t i on and

mate r i a l i za t i on i n a rch i tec tu re , AMIT I n te rna t iona l Jou rna l fo r A rch i tec tu re and

Modern I n fo rma t ion Techno log ies , Vo l . 4 No. 03/2010.

Menges , A . 2008. I n teg ra l fo rma t ion and mate r i a l i sa t i on : Computa t i ona l fo rm and

mate r i a l ges ta l t . I n Manu fac tu r i ng Mate r i a l E f fec ts : Re th ink ing Des ign and Mak ing

i n A rch i tec tu re , eds . B . Ko la rev ic and K . K l i nge r, 195 – 210. New York : Rou t l edge .

Te rz id i s , K . 2006. A lgo r i t hm ic a rch i tec tu re . Ox fo rd : A rch i tec tu ra l P ress .

Wagen füh r, A . 2008. D ie s t ruk tu re l l e An iso t rop ie von Ho l z a l s Chance fü r techn ische

I nnova t ionen . S tu t tga r t : H i r ze l Ve r l ag .

Wagen füh r, R . 1999. Ana tom ie des Ho l zes . Le in fe lden-Ech te rd ingen : D rw Ve r l ag .

computation, formation and materiality