sagrada família rosassa: global computer- aided dialogue

2001: ACADIA76

Sagrada Família Rosassa: Global Computer-aided Dialogue between Designer andCraftsperson (Overcoming Differences in Age,Time and Distance)Mark BurryDeakin University, Australia

Jane BurryDeakin University, Australia

Jordi FaulíSagrada Família Church, Barcelona

AbstractThe rose window (‘rosassa’ in Catalan) recently completed between the twogroups of towers that make up the Passion Façade of Gaudí’s Sagrada FamíliaChurch in Barcelona measures eight metres wide and thirty-five metres in height[Figure 1]. There were four phases to the design based in three distinct geo-graphical locations. The design was undertaken on site, design description inAustralia some eighteen thousand kilometres distant, stone-cutting a thou-sand kilometres distant in Galicia, with the completion of the window in March2001. The entire undertaking was achieved within a timeframe of fifteen monthsfrom the first design sketch. Within this relatively short period, the entireteam achieved a new marriage between architecture and construction, a broaderrelationship between time-honoured craft technique with high technology, andevidence of leading the way in trans-global collaboration via the Internet.Together the various members of the project team combined to demonstratethat the technical office on site at the Sagrada Família Church now has thecapacity to use ‘just-in-time’ project management in order to increase effi-ciency. The processes and dialogues developed help transcend the tyranny ofdistance, the difficult relationship between traditional craft based techniqueand innovative digitally enhanced production methods, and the three genera-tional age differences between the youngest and more senior team members.

KeywordsDigital Practice, Global Collaboration, Rapid Prototyping

SimulationSection 1

2001: ACADIA 77

1 IntroductionJust-in-time construction is a relatively new pro-cess where construction takes place for one partof the project while others parts are still havingthe detailed design finalised, and is a recent ini-tiative taken by the Junta Constructora as one moremeans to advance the completion of the churchby accelerating the building process. Of course,‘just-in-time’ has always been the design-buildmethod that has characterised the long gestationof the Sagrada Família Church, even duringGaudí’s time, but its application in such a fast-track way to the Passion Façade rose window us-ing the latest digital design and communicationtechnology was a first for the project, and prob-ably rates as a first for any project of this nature.It was made possible through the project man-agement of Architect Jordi Faulí working closelywith the Coordinating Architect Jordi Bonet, andConsultant Architect professor Mark Burry andhis team at Deakin University. As such, and as

part of an ongoing process to development tech-niques to increase the productivity of the designand construction of this internationally renownedyet unfinished project, the work in progress re-ported here yields significant insights into the‘Reinventing design, construction and operationwork processes’ section for Acadia 2001.

2 ScopeThe project was in four distinct phases startingwith the overall design. Mark Burry worked onsite in Barcelona using a high-end computer for aperiod of six weeks at the beginning of 2000.Using a particular software approach commonlyreferred to as ‘parametric design’, a flexible 3Dcomputer model was made of the whole windowassembly; by flexible we refer to a model whosecharacteristics could be adjusted to suit knownenvironmental conditions. For the rose windowthe team was departing for the first time fromdetailed models of windows prepared by Gaudíhimself - there is no surviving concrete proposalby Gaudí for this element, and only the vaguestof ideas can be captured in his surviving drawingof the Passion Façade [Figure 2]. We do havetwo important clues to his intentions, however,the first being the existence of the Nativity Façaderose window on the opposite transept, and thesecond being the lineage of Gaudí’s 1:10 scale plas-ter models for the lower and upper side aisle win-dows culminating with the central nave windowin terms of design complexity (Burry, 1993). Thislineage shows a developing sophistication withinthe articulation between the adjacent hyperbolicsurfaces and accordingly we have assumed thatthe rose window would highlight the particulari-ties of all three antecedents (Burry et al, 1996).

3 Parametric modelling1:25 scale gypsum plaster studies of a proposalmade by the architects Puig i Boada and BonetGarí during the 1970s [Figures 3 & 4] proved tobe an excellent starting point. Their overall com-position was updated through the application ofthe Coordinating Architect and Director Sr JordiBonet’s understanding of Gaudí’s system of pro-portions (Bonet, 2000). Mark Burry was able touse this model as a first basis for the final design,and by using the system of parametric design, thewhole composition could be altered at will. AnFigure 1. Position of the rose window from the interior

Mark BurryJane BurryJordi Faulí

Sagrada Família Rosassa: Global Computer-aidedDialogue between Designer and Craftsperson

(Overcoming Differences in Age, Time and Distance)

2001: ACADIA78

example of this variability is the change of one orall of the mathematical constants that govern thecharacteristics of the component hyperbolic sur-faces [Figure 5]. A particular surface, for instance,can be made more full (deeper) through chang-ing the mathematical characteristics of the sur-face via the computer software, thereby affectingthe nature and position of the intersections ofadjacent surfaces [Figure 6]. The position of anopening, or the size of the opening itself can bechanged at will with the 3D computer model au-tomatically reforming itself without the need torebuild the digital model through erasure andsubsequent redrawing. In the case of this win-dow, there are 3,800 events in the parametricmodelling history; that is to say, for any charac-teristic of the model to be changed, 3,800 eventsneed to be checked and updated before the ge-ometry of the model can be regenerated.

Parametric design or ‘associative geometry’ is anapproach to design that has been on the palette

Figure 2. Original drawing by Gaudí – ellipse marks the posi-tion of the rose window

Figure 3. Provisional 1:25 gypsum plaster model (1970s) Figure 4. Provisional 1:25 gypsum plaster model (1970s)

SimulationSection 1

2001: ACADIA 79

of possibilities for many of the 3D design profes-sionals but has proved elusive for architects todate. There are very few examples where the op-portunities for reconfigurable modelling based onassociated geometry, which is a systemisation ofgeometrical dependencies whereby the designercan revisit earlier decisions and rework the modelwithout the usual erasure and redrafting. Theonly example known to the authors is the roofstructure for the Eurostar section of the Water-loo Station roof in London. The reason for itslack of uptake has been the prohibitive pricebracket for architects, and possibly a reaction toany ‘design tool’ that encourages reiterative de-sign processes. The second obstacle cannot beidentified – it is more presumed and cannot becontested until sufficient architects have assessed

the opportunities through practice rather thanpresentation. The first obstacle, however, can bechallenged, and at least one major AEC softwaremanufacturer (Bentley) is well advanced with amore affordable associative geometry modeller.

Naturally, Gaudí had absolutely no interest inanything as esoteric as computers and software inhis day. The lineage of craftspeople unbrokenboth during his time and in the years since hisdeath in 1926 have alluded to his own modus oper-andi; although effective in his day, neither thennor in our time can a process based on reiterativemodelling be anything less than lengthy [Figure7]

We sense a similarity between this particular modusoperandi using powerful digital aids and the hap-tic iterative design modelling used by Gaudí dur-ing the pre-digital era.

As complete as they may purport to be, high-endsoftware packages still require somecustomisation, however. In the case of hyperbo-loids of revolution of one sheet, for example, thesecannot be conveniently called-up from a libraryof geometries. Nevertheless, the basis by whichuseful geometries are sought using a reiterativedesign process whether it be gypsum based ordigitally assisted, achieving greater efficienciesallows for more option-testing in the availabletime. To model a hyperboloid of revolution, testit, and refine it subsequently takes an inordinateamount of extra time if it cannot be experimentedwith through parametric variation on the fly. Inthe case of CADDS5, we resorted to an amalgamof mathematical calculation via MS Excel to de-rive the appropriate geometries, and MS Wordvia macros to convert that information into ex-ecutable scripts for CADDS5 [Figures 9 and 10].To the computer-using aficionado, reverting tostandard desktop in order to produce the execut-able scripts probably seems top-heavy. On thebasis of ‘just getting on with the job’, such pur-ism is interesting, but not especially relevant.

After three months the design team were satis-fied with the overall design composition for thewindow; the exterior was fully resolved while theinterior had aspects that had to wait until certainmeasurements were made on site during construc-

Figure 5. Parametric model of window (CADDS5)

Figure 6. Parametric model of window (CADDS5)




2001: ACADIA80

tion before being finalised. Again, the use of para-metric design technology allowed the easy absorp-tion of the new information as it became avail-able with a minimum of repeated work [Figures11 – 13].

4 Rapid prototypingDuring the creation of the overall digital designmodel during the first month, the stonemasonManuel Mallo made several visits from Galicia tothe onsite design office in Barcelona to discussthe project in detail – his quarry and stonemason’syard are approximately 1,200 kilometres distant,and Sr Mallo himself is in his seventy sixth year.This was the first time the team had attempted towork directly from digital models to manually-based stonecutting without preparing an interme-diate scaled gypsum plaster maquette. As a result

Figure 7. Gypsum plaster model showing Gaudí’s method of re-iterative form finding. Here we can see a number of intersectinghyperboloids of revolution of one sheet, all of which would havebeen derived from drawn sections (hyperbolas) shown in figure 8.

Figure 8. Graphical derivation of hyperboloids of revolution ofone sheet

Figure 9. MS Excel is used to derive the characteristics for par-ticular geometries.

Figure 10. The calculations performed in MS Excel are thenconverted into tables that are then automatically processed intoexecutable scripts to ‘drive’ the parametric model. The whole pro-cess of trying different variables for the geometry, producing ascript through MS Word macros and running it within CADDS5is a matter of seconds.

SimulationSection 1

2001: ACADIA 81

Figure 11. Computer render of exterior (Rhino3D)

Figure 12. Computer render of exterior (Rhino3D)

the design-sculpting team had to invent an en-tirely new way of working together, not leastthrough all three groups being based at inconve-nient distances from each other. Through SrMallo’s active involvement, we were able to de-velop the means that together could resolve thewindow efficiently into component pieces for cut-ting and shaping in his workshops on the otherside of Spain. The issue here was a maximumweight to suit the cranes at both end of the op-eration, the sizes appropriate for trucking acrossSpain, and the manoeuvrability of the pieces whensuspended at the working zone many tens ofmetres above ground level.

Included in this process was the Sagrada FamíliaChurch’s first committed (as opposed to scepti-cal) adventure into rapid prototyping using aThemojet ™ 3D printer, which produced exquis-itely detailed, accurate scaled versions of each ofthe stone components that typically weigh sev-eral tonnes [Figure 14].

Manuel Mallo had two observations to make dur-ing this critical phase. The first point was thatthe experimentation into the use of new technolo-gies such as the 3D printer took longer than thecritical path allowed and is best undertaken in aless demanding context. The second point wasthat his own manual method of rapid prototypingusing polystyrene is in fact more rapid albeit lessprecise than automated rapid prototyping, withthe advantage that the model maker understands

Figure 13. Wireframe of window for resolving the computer wholeinto individual stone pieces (Rhino3D)




2001: ACADIA82

more about the task in hand through the model-ling than would occur with a machine made model[Figures 15 & 16].

5 Templating, stone cutting and buildingOf particular concern was the nature of the infor-mation with which Sr Mallo could direct his teamof stonemasons. The lower part of the windowwent into production in March 2000, during thethird month from the commencement of theproject. This, a relatively uncomplicated part ofthe window, was an ideal testing bed for our newapproach to design representation, rapidprototyping and production documentation. Inlate April, Architects Jordi Faulí and Mark Burryvisited Sr Mallo’s atelier near Lugo for an illumi-nating discussion on the lessons learned fromcutting the lower window components for this firstphase.

By mid year, while the lower quarter of the win-dow was being constructed on site, the secondquarter was being cut in Galicia, the third quar-ter being made into templates in Australia to guidethe stonecutting, while the top quarter was stillhaving the design refined in the collaborationbetween the Sagrada Família design office and theDeakin University team in Australia. This wasthe first instance of ‘just-in-time’ constructionbeing put to effect at the Sagrada Família Church.

The window is more complex in design with in-creasing height, and the increasing complexity indesign sponsored increased complexity for thetemplate production leading to a second visit toGalicia in late summer. At this point it had be-come necessary for the stonemasons to first pre-pare full-scale prototypes for each stone pieceusing expanded polystyrene. These facsimiles

Figure 14. Rapid prototype of individual stone piece (3D Sys-tems)

Figure 15. Manuel (stonemason) and Jordi Faulí with hand-made expanded polystyrene rapid prototype. Figure 16. Manuel Mallo’s hand made rapid prototype.

SimulationSection 1

2001: ACADIA 83

have the advantage of weighing only a fraction ofthe actual pieces, some of which weigh severaltonnes, and the expanded polystyrene is certainlya great time-saving improvement on the gypsumplaster Gaudí was obliged to use for the equiva-lent process when building Casa Milá almost onehundred years earlier (Bassegoda, 1989).

Nearly eight hundred full-size DIN A0 templateswere required to guide the stonemason’s hand.These drawings were not based on any precedentthis being the first time Gaudí’s use of secondorder geometry (ruled surfaces) had been appliedto the project in this way. A combination of ren-dered images of each piece [Figure 17] and asso-ciated full-size templates [Figures 18 & 19] weresent from Australia to Barcelona at the end of eachday, printed and proofed there before being for-warded to Galicia by courier the same day. The system developed relied on line colour to repre-

sent the nature of each line (green lines beingsurface generatrices (therefore straight lines),black a template boundary between adjacentpieces, orange a curve etcetera). The drawingswere supported by surface contours at separationsof 10 cm (pale blue) with coordinates at criticalnodes linking the orthographic projections of eachpiece to a common datum for each piece.

Manuel Mallo had introduced a number of inno-vations to his stone cutting procedures, principalof which was a semi automated cutting of ruledsurfaces using an abrading wire. The piece ofstone could be steered by an operative at each endof the hydraulically assisted controlling devicewho would each follow a path described by thetemplates. In this way he was able to cut warpedsurfaces to within a centimetre of the finishedoutcome, the finishing achieved relatively quicklyby hand [Figures 20 – 21].

With the exception of a brief second visit by MarkBurry to Talleres Mallo near Lugo in late August,during the period June – December the Barcelona– Deakin University collaboration was conductedentirely via the Internet. Of note was the useful-ness of the parametric design modelling in allow-ing for relatively easy adjustment of the base modelto match the real circumstances of the adjacentwalls. Maintained in a fairly ‘loose’ configura-tion, specific onsite information that could onlybe practically measured once the scaffolding had

Figure 17. Computer rendering of typical individual stone piece

Figure 18. Typical A0 template for semi-automated stone cut-ting (AutoCAD 2000)

Figure 19. Detail of A0 template showing system of coordinatesand contours for spatial information




2001: ACADIA84

been put in place after the lower part of the wallconstructed could be factored in and the para-metric model retuned to match with exacting pre-cision [Figures 23 & 24]. Following this process,and without exception, the large number of A0templates sent to Lugo from Australia via theSagrada Família site office were put to immediateuse; none required any revision. Equally withoutexception, all the stone pieces from Galicia werefitted together first time on site making the wholewindow without any significant on-sitestonecutting or revision.

6 Concluding commentsThe singular trials of the one-off nature of build-ing construction can hardly be better tested thanthrough a Gaudí project. This short paper con-ceals many intriguing aspects that are publishedelsewhere. Not least is the fact that Gaudí in-vented a kind of codex by which the building’scompletion could be managed by others follow-

Figure 21. Lower four courses at the stonemason’s yarding his demise, hardly expected at the time dur-ing his advancing years (he was killed by a tram atthe age of seventy-four). He nevertheless antici-pated that the project would not be completed inhis lifetime. The codex - a geometry based ratio-nality – lends itself to collaboration between di-verse trades and professions not necessarily tiedto one location. This latest application of digitaltechnology to manually based craft represents aconsiderable advance towards more efficient andcost-effective construction of this nature.

The experience of introducing rapid prototypingled, however, to two important revelations. Al-though subsequently established as an invaluablevisualisation aid on site, trying to apply such aradical shift in process during a just-in-time ‘ex-periment’ was ill-conceived revealing, it seems,that such incursions are better tested in the aca-demic laboratory than within a tight critical pathin a real-world situation. The second revelationwas the further substantiation of the premise that

Figure 20. Manuel Mallo next to typical stone element cut bymachine (surfaces ‘a’) and finished by hand (surfaces ‘b’)

Figure 22. Individual stone piece on site

SimulationSection 1

2001: ACADIA 85

Figure 24. Close-up photograph showing of the precision givenusing parametric design (refer to figure 6)

Figure 25. partially constructed window (finished and glazed atthe commencement of 2001, but still site-protected and thereforenot available for full-viewing at the time of publication)

Figure 23. Example of the precision given using parametric de-sign (refer to figure 6)

the innate precision of the digital environment issometimes costly within the relatively crude tol-erances of the building industry, and that hapticengagement has its own rewards within a processultimately as hand-driven as the craft of cuttingstone.

AcknowledgmentsThe authors gratefully acknowledge the JuntaConstructora of the Sagrada Família Church andthe Australian Research Council for funding thisresearch over a three year period. The researchreported here also acknowledges the support andexpertise of Patrick Dunne from 3Dsystems (UK)for the ThermoJet™ 3D printing work for therose window.

ReferencesBassegoda I Nonell, Joan; La Pedrera de Gaudí, Fundacio

Caixa de Catalunya, 1989

Bonet, J, (Burry, M.C. trans), L’Ultím Gaudí, EncyclopaediaCatalana, Barcelona Spain, May 2000

Burry, M.C., The Expiatory Church of the Sagrada Família,Phaidon Press Ltd, London UK, 1993

Burry, Coll, Goméz, Melero, La Sagrada Família: de Gaudíal CAD, Ediciones Universitat Politècnica de Catalunya,Barcelona, Spain, 1996




2001: ACADIA86

sagrada família rosassa: global computer- aided dialogue

Documents