building performance modeling in non-simplified architectural design; max c. doelling, farshad...

7/22/2019 Building Performance Modeling in Non-Simplified Architectural Design; Max C. Doelling, Farshad Nasrollahi.

1/10

97CAAD curriculum -Volume 1 - eCAADe 30 |

INTRODUCTIONDigital, parametric model-based design workows

oer many opportunities to integrate perormance

simulation into the architectural design process, but

as a relatively novel practice, no proven set o de-

sign methods or cognitive ramework has yet been

established. Many traditional simulation classes

consider simplied design parameters and produce

results that stream towards clear perormance indi-

cators. While entirely appropriate, and possibly even

reects a large aspect o the built environments or-

mal reality, an increasing tendency exists to strive

towards orms that are not intended as mere aes-

thetic experiments but to enrich the lives o inhabit-ants through enhanced comort. In this context, our

ongoing seminar Parametric Design investigates

the integration o multiple building perormance

simulation techniques into the early stages o archi-

tectural design. Master o Architecture students with

minimal or no knowledge o building perormance

simulation are tasked with expressing a unction-

ally diverse spatial programme, using daylighting

and thermal assessment tools as continuous design

decision benchmarks. One o three sites (Berlin: Ger-

many, Hashtgerd: Iran, Ft. Lauderdale: Florida, USA)

has to be chosen, yielding designs specic to the lo-

cal climate but related through their shared design

brie. Basic lectures on sustainable building andsimulation principles are given, while workshops

Building Perormance Modeling in NonSimplifed

Architectural Design

Procedural and cognitive challenges in education

Max Doelling1, Farshad Nasrollahi

2

1,2Technische Universitt Berlin, Germany

1http://spacesustainers.org,

2http://www.enef.co

[email protected],

[email protected]

Abstract. The building technology class Parametric Design simultaneously teaches

thermal and daylight performance simulation to novice users, usually Master of

Architecture students. Own buildings are created, analysed and geometrically modiedduring the design process, resulting in structures that are energetically pre-optimized. It

is shown that energy demand and daylight utilization can be signicantly improved while

taking into account formal considerations. Departing from a design process model that

gives preference to either engineering or design thinking, multi-modal decision-making is

diagnosed to be mediated by hybrid or multivalent representations, necessitating a shift

in how inter-domain design knowledge ows might be understood. Opposed to purely

linear or iterative process assumptions, a uent state model of interconnected domains of

analytic inquiry is proposed.

Keywords. Sustainable design; daylight simulation; thermal simulation; architecturaleducation; design epistemology.


2/10

98 | eCAADe 30- Volume 1 - CAAD curriculum

introduce students to the interlinked usage o DIVA

(Jakubiec and Reinhart, 2011), a daylight simulation

plugin or Rhinoceros3d, DesignBuilder, an interace

or the simulation engine EnergyPlus, Rhinoceros3d,

a NURBS modeler, and Grasshopper3d, a parametric

geometry tool.

PRECEDENTS IN DESIGN-SIMULATIONINTEGRATIONVarious dierent models o building perormance

simulation classes are described in the literature, as

are approaches that deal with integrating simula-

tion into the architectural design process in general.

The ollowing selection is not intended as a com-

prehensive classication o previous studies, butinstead serves to position our own endeavor within

this developing eld.

Many simulation classes cater to architecture

students and contain a design component, the

analysis o which can then also be related to tool use

considerations (Palme, 2011). Alternatively, uncon-

strained architectural design activity is requently

not part o a class (Strand, Liesen and Witte, 2004;

Madsen and Osterhaus, 2005). Epistemologicalworkow considerations are discussed rom various

angles, usually contrasting engineering and design

working methods, or attempt to establish interme-

diate ground (Batty and Swann, 1997; Hetherington

et al., 2011; Venancio et al., 2011). Most case studies

acknowledge the importance o early-stage archi-

tectural energy optimization through design, yet

our review indicates that it is the norm or only one

simulation domain to be detailedly taught per class,

unlike in our own, which introduces both daylight

and thermal simulations in an integrated manner.

In this paper, we touch on tool use implications, but

assume the chosen sotware applications to be reli-

ably useable in the design process due to advanced

interaces, precise results display and their use o the

validated simulation engines EnergyPlus and Radi-

ance. In essence, we attempt to understand possible

modes o simulation-assisted cross-domain deci-

sion-making by architectural designers perormedrom the very rst creation steps to the end o the

schematic design phase, since in this time rame

undamental, difcult to reverse choices aecting

orm and energy use are made (Brown and DeKay,

2001). No normative workow recommendations

will be stated; instead, analyzing various process

representations in conjunction with describing deci-

sion chains will lead us to an alternative integrated

design process model.

RESEARCH QUESTIONSThe ollowing sections introduce the curriculum em-

ployed by us, present two class results rom summer

2011 and relate them to the challenge o integrating

building simulation into early-stage architectural

design. The guiding research questions are:1. Are simulation activities easily eective in de-

creasing a designs primary energy demand i

they are to be positively correlated with mak-

ing desired ormal decisions, in a process that

acknowledges unctional and geometric com-

plexity?

2. How are design decisions made in a multi-rep-

resentational domain that includes parametric

perormance models?3. What consequences might the results and the

modes o their making have or architectural

education and design theory?

CLASS ORGANIZATIONThe seminar investigates design-simulation process

interaction by posing a real-world problem. There

are no rules concerning the building shape, albeit

we ask students to consider the task realistic in the

sense o a limited budget and apparent constructa-

bility o the 804 m2

community center. Spaces are

a mix o ofces, seminar rooms and a small audito-

rium. By having students design in dierent climate

zones, they experience how buildings that share the

same design brie are morphologically inuenced

by adapting to the local climate.

CURRICULUM, ASSIGNMENTS AND

DESIGN OBSERVATIONSThe course assignments are modeled ater the hier-


3/10


archy traditionally ound in design studios, with en-

ergy optimization primarily to be achieved through

architectural instead o technological means. Hence,

ormal and perormance decisions are closely re-

lated (Nasrollahi, 2009). Our ollowing process narra-

tive is a temporally linear approximation o groups

design thinking at various advancing stages, as in-

terpreted by the authors based on tutoring, results

data evaluation and representation analysis.

Heuristic and initial simulation phaseAssignment 01 asked students to start design work

and to especially consider early stage perormance

rules o thumb, requiring them to document key de-

sign concepts relating to the intended environmen-tal perormance in principle sketches that should

relate to the local climate. Group 01 chose Ft. Lau-

derdale as the project site; group 02, chose Hashtg-

erd, Iran. The climates o both sites are very dierent:

Floridas low latitude, low elevation and proximity to

the Gul o Mexico lend it year-long high tempera-

tures, mostly uncomortable in summer due to high

relative humidity, while Hashtgerds more northern

latitude, higher elevation and greater distance romthe ocean yield a continental climate with both

summer and winter discomort extremes. The goal

o perormance design strategies was to minimize

the primary energy demand o heating, cooling and

lighting equipment required to achieve thermal and

visual comort.

In the initial phase, a massing approach (gure

1) and response to site conditions was dened, with

most groups departing rom and modiying the ba-

sic principles thus discovered throughout the class.

Design rules o thumb were recommended to be

ollowed rom various publications (e.g. Brown and

DeKay, 2001) and Climate Consultant, a climate anal-

ysis package. The very earliest design stage, articu-

lated by sketches (gure 2) and arrays o incomplete

models, eatured too little inormation to enable

simulations that require dened geometries. Only at

its end was a base layout established, on which rst

analyses were perormed. This marked the transitionrom purely heuristic to partially evidence-based

modes o thinking.

Assignment 02a dealt with running climate-based

daylight simulations (Reinhart, Mardaljevic and Rog-

ers, 2006) and a seasonal cumulative solar irradiation

analysis, achieved with DIVA. The reason or consid-

ering insolation images and daylighting rst was our

intention to have students visually experience solar

gains, with the hope that they would tweak their

assumptions on solar geometry and arrive at an im-

proved building layout beore constructing thermal

simulation models.

Group 01 correctly identied horizontal as well

as east and west acing suraces as major receivers

o insolation (gure 3). This was in part caused by

the massing strategy chosen due to wind patterns

and spatial organization considerations, which werealso related to assumed daylighting benets tested

Figure 2

Design Development Sketches

(Plans roughly oriented north),

USA.

Figure 3

Summer / Winter Irradia-

tion Images (South to let o

rame), USA.

Figure 1

Natural Ventilation and Mass-

ing Sketch, Florida Design.


4/10


through simulations and ound to be promising

(gure 4). The design process in all coming phases

then evolved towards a systematic evaluation o di-

erent acade structures, many o which were not

iteratively related but ormal experiments and per-

ormance assessments in unison. Aesthetic require-

ments o retaining vertical ns to visually balance

the horizontal massing scheme were expressed by

the group throughout the class, setting a ormal pa-

rameter space within which most explorations were

achieved. The authors ound that this happened

in the case o most groups; the nal solutions re-

quently showed an expression o ideas developed

during the heuristic design development phase.

The Iran team ran site-level irradiation simula-tions via Ecotect and chose a site patch with maxi-

mum insolation to receive their design, intended as

a compact volume tilted towards lower sun angles

(gure 5). Despite at rst glance promising, it later

became apparent that this systematic initial ap-

proach yields no guarantees that building peror-

mance will actually be superior to rule o thumb

only approaches, since when site-level analyses

are perormed without preconceived ideas on thestructure to be designed, no relationship between

measured site phenomena and building geometry

yet exists.

Interior spaces were arranged into a dense lay-

out situated under a slanted roo perorated by sky-

lights. The handling o these apertures was the key

geometric element aecting design perormance;

they developed rom simple horizontal openings

to complex solar scoops, their behavior parametri-

cally dened by exible Grasshopper3d-denitions.

Insolation analysis perormed on various scoop tilts

resulted in the group choosing an angle that caused

greater gains in winter and relative prevention o

direct sunlight penetration in summer (gure 6). In

that sense, the irradiation images played a greater

role in meshing thermal optimizations with ormal

considerations, and thus acted more as a useul

tool than they did or group 01, who argued rom

a dierent set o constraints, especially wind pat-terns and projected daylight demand. Group 02s

approach can hence be understood as having been

more driven by thermal perormance concerns; the

observed useul daylight utilization o the rst itera-

tion was indeed sub-par (gure 7).

Figure 4

Variant 01 (Florida), Useul

Daylight Illuminance, 100

2000 lux, % o occupied hours,

xed louvers.

Figure 5

Site Irradiation Analysis & Vol-

ume Derivation, Iran Design.

Figure 6

Summer / Winter Irradia-

tion Images (North to top o

rame), Iran.

Figure 7

Variant 01 (Iran), Useul Day-

light Illuminance, 100 2000

lux, % o occupied hours, no

shading.


5/10


Detailed simulation phaseAssignments 02b and 03 required students to adapt

their designs through DesignBuilder (DB) and ur-

ther DIVA simulations. Alternate massing strategies

had to be considered in step 02b; in assignment

03, the best perorming massing variant in terms

o primary energy demand and daylight utilization

was to be chosen and several design actors system-

atically varied to arrive at a nal proposal, its energy

and daylight perormance to be ully analyzed. By

keeping simulated physical building materials, occu-

pancy inormation and assumed best-practice HVAC

templates constant throughout the class and con-

centrating on changes on the level o orientation,

massing, glazing ratios and xed shading geom-etries, the direct inuence o orm on perormance

was studied; yet in practice, initial decisions usually

overrode the possibility o undamental changes.

Most groups ound it hard to divorce their thinking

rom the version already created and to dene an al-

ternate massing scheme.

Expressing the desire to retain the initial design,

group 01 departed rom an unshaded base design

(gure 8) and especially studied the eects o side-n shading geometries and ventilated double-roo

structures (gure 10) on total energy demand, re-

ducing it by 30%. Gross daylight utilization was im-

proved 15% by using light shelves and modiying

n spacing (gure 11). Light shelves were used or

all but the North orientations and additionally acted

as overhangs (also see gure 20). Simulation results

are summarized in gure 15, clearly showing an in-

crease in overall design perormance. The required

alternate volumetric scheme o variant 02b (gure 9)

did not have an impact on subsequent design deci-

sions, possibly due to its negligible perormance im-

provement and seemingly improvised layout.

Group 02 did not produce an alternate mass-

ing scheme, but instead ocused on the spacing,

arrangement and glazing area o the skylights, also

starting rom a base design (gure 12). The number

o aperture rows in the nal iteration was reduced

by three and the total glazing area more than halved(gure 13), which lessened total energy demand

Figure 8

DB Model, Variant 01, USA.

Figure 9

DB Model, Variant 02b.

Figure 10

DB Model, Variant 03.


6/10


by 25% and almost doubled useul daylight utiliza-

tion (gure 14). An increase in skylight row spacing

meant a reduction o winter overshadowing eects;

this change was stimulated by knowledge gained

rom the previous irradiation image analysis.

There was considerable geometric drit be-

tween the individual design and simulation models,

as well as strong abstractions present in the ther-

mal models. The most pronounced difculty lay in

how to port the light scoop geometries between

daylighting and thermal models; this was solved by

synchronizing the glazing area and building cus-

tom overhangs in DesignBuilder, which imitated

the scoop tilt as used in the Grasshopper denition.

Opposed to the Florida team, who perormed inter-mediate daylighting tests on singular geometric ex-

pressions and generally kept DB and Rhino models

parallelized, group 02 used several thermal geom-

etry variants independently. Two series o models

with a stepped decrease in scoop glazing area were

compared and the results ed back into the original

parametric geometry denition (gure 16). As such,

an iterative workow was contained within a ormal

parameter space, which was itsel dynamically en-coded and eventually updated to reect the nal

analysis step.

Naturally, the groups results in both simulation

domains could be improved, yet by limiting material

choices to elucidate the eects o orm and being con-

strained by what simulation novices can accomplish in

a single semester, more detailed optimizations had to

be deerred. Furthermore, the nal absolute numbers

are not the primary result; rather, it is the comparative

evaluation o geometric inuences on perormance

that makes up the value o the simulations. More de-

veloped models would likely yield dierent results,

since more precise interaction eects o daylight qual-

ity, which is not readily described by bulk UDI values,

and window shading would modiy design peror-

mance, as would urther thermal comort and natural

ventilation considerations. Group 01 again improved

design perormance leading up to the rapid proto-

typing stage, during which nal models were printedwith daylight metrics embedded (gure 17).

Figure 11

Variant 03 (Florida), Useul

Daylight Illuminance, 100

2000 lux, % o occ. hours,

ns only.

Figure 12

DB Model, Variant 01, Iran.

Figure 13

DB Model, Variant 03.


7/10


From a process perspective, workows that meshed

iterative tests with the concurrent exploration o

other related but singular design variants appeared

as the norm; while we provided extensive instruc-

tions on how, in our opinion, to best structure an

analysis workow, the ot-articulated conict be-

tween design and engineering thinking came into

play, but without inhibiting a measurable decrease

in energy demand and a general increase in daylight

utilization. Group 02s systematic, iterative approach

did not automatically produce a design that per-

ormed better than the Florida teams building.

DISCUSSION AND CONCLUSIONS

Most students accomplished a positive interplayo geometry and perormance actors. The easibil-

ity o a mixed design-simulation process in achiev-

ing efciency improvements was demonstrated,

however it is not only through raw data that such a

practice must be evaluated. More than the sum o

its parts, it becomes an activity o mediation, com-

plicating both epistemes by collapsing them into

the same space o thinking and evaluating. I care-

ully managed, and to begin answering the rstresearch question, quantitative improvements can

be achieved in an integrated manner and through

iterative evaluations accompany and even inspire

ormal experimentation. On the other hand, synergy

breakdowns can also occur, experienced by a mi-

nority o groups that ailed to connect the domain

o analysis with the domain o creative production,

usually caused by a lack o basic building science

knowledge. For i epistemes are to intersect, they

need to be at least rudimentary developed, inde-

pendent o how knowledge is actually produced in

science versus design methodologies.

Apart rom concerns o principle easibility, we

implied the question o whether a combined de-

sign-simulation process would easily increase per-

ormance. This can only be answered in conjunction

with the core question o how design decisions were

made in general and specic instances. Since design

is oten understood as a goal-oriented activity, deci-sions cannot be evaluated in isolation, but need to

Figure 14

Variant 03 (Iran), UDI 100

2000 lux, % occ. hours, no

shading.

Figure 15

Iteration Perormances.

Figure 16

Parametric Roo Scoop Geom-

etry o Iran Design.


8/10


be seen in relation to the perceived whole. Design

intent is requently articulated in a nonlinear and in-

tuitive manner striving towards synthesis by accom-

modating possibly clashing goals, and thus bears

conict potential with rationalist engineering proce-

dures. Intent encapsulates the always current total-

ity o ideas on how a building should be (N, gure

18), but due to its complexity and intersubjectivity

has no holistic representation. It includes all design

assumptions, also the ones related to perormance,

and at any given moment can be understood as a

uent total state o ambivalent interconnections,

exemplied by Christopher Alexanders chart o

design actor interdependencies (gure 19). Alexan-

ders chart predates the availability o digital designand simulation models, but nonetheless deals with

material and social perormance interdependen-

cies that orm a wicked problem (Rittel and Webber,

1973).

Architects, especially since their separation

rom manual construction activities (Davis, 2000),

have developed a tradition o dealing with problem

subset permutations o dierent domains that still

relate to the same object, e.g., how to marry struc-tural requirements with space ow demands. These

dierent subsets are traditionally encoded by a mul-

titude o space-related drawings and models that

reer to the same object but are still unique episte-

mes. As process models, they can act as machines

or thinking (Smith, 2004) and enable associative ar-

tistic leaps. Given that in our case study most projec-

tive representations and perormance datasets were

derived rom multiple digital models, strong clues

exist that model amilies may in act be used by ar-

chitects within a contemporary continuation o said

historic ramework, which has been perpetuated by

educational design studio practices situated in the

lineage o Modernism. Yet since its heyday, devel-

opments in simulation and its space-related repre-

sentation have moved numeric evaluations much

closer into architectural planning practice. Still in

an apparent conict with design nature, analysis

necessitates clear steps in a rational procedure andrelies on steady benchmarks during simulation, oth-

Figure 17

Printed Daylight Model, UDI

100 - 2000 embedded, USA

Design.

Figure 18

Domains o Inquisition /

Representation in Design

Synthesis.

Figure 19

Field o Design Factor Interde-pendencies (Chermayef and

Alexander, 1963).

Figure 20

Principle Perormance Section:

Multivalent Representation,

USA.


9/10


erwise invalidating comparisons. Yet creative reality

is prone to upheavals questioning the very stability

o the contained analysis paradigms. How, then, was

their interplay managed?

The design observations show that a combina-

tion o heuristics, to establish an initial ormal seed,

and iterative schemes was usually employed, the

latter o which predominantly revolved around per-ormance evaluations o building components and

were strongly related to pregured intent; as such,

they were encapsulated by and inseparable rom

the heuristic context. Models that dealt with dier-

ent design aspects drited apart, were abstracted

to explore isolated perormance behaviors and

later synchronized with master design models, as

shown by group 02. Parametric encoding can be

understood as a process analogue to the creationo myriad manual test models and was used to simi-

lar design renement ends. Other groups exhibited

related behavior; a multiplicity o independent but

related digital models was used to generate analytic,

orm-related and, most importantly, hybrid or mul-

tivalent representations concerned with the orm-

perormance interace and acting as design cata-

lysts. We observed that most benecial perormance

decisions were made when students either achieved

a parallel presence o design intent across multiple

representations belonging to dierent domains o

inquiry, or created multivalent representations that

directly combined validated assumptions rom mul-

tiple domains. To extrapolate a model:

Individual domain-specic types o knowledge

(An

etc., gure 18) are synthesized by utilizing the

semiotic exibility their multivalent representations

(e.g. derived rom digital models) enable, and thus

continuously update global design intent (N, gure18). In return, the eld o intent, newly enriched with

additional cross-domain knowledge, permanently

inuences the originally contributing domains,

orming a nonlinear knowledge ow ramework

that relies less on direct hybridization o design and

engineering methods, but instead draws potential

rom the synergistic possibilities rooted in the multi-valence o their respective models representability.

This model neither invalidates the presence o

engineering procedures nor the validity o grown

design methods, but in part shits the discourse

onto the level o understanding the mediating

role o multivalent representations (e.g. gures 20

& 21), which by virtue o their properties encode

quantitative descriptors spatially, relate orm to

projected perormance and should be regarded asarticulating one possible state o synthesis among

many. The shown sections, daylight plans, radiation

images and printed daylight models all partially

ulll these requirements. In a process model that

is perceived as a eld o possibilities managed by

denitions achieved through representations, all

contributing domains constantly interact. Repre-

sentations stimulate processes, can be their result

and by eedback eects cause shits in their re-spective knowledge source domains; as an exam-

ple, we ound that by running many consecutive

simulations, students became increasingly good

at without urther tests predicting how glazing ra-

tio changes would impact combined thermal and

daylighting perormance. Yet in order to establish

that relationship, it in most cases had to be previ-

ously encoded in either conceptual drawings or

numerical representations that clearly meshed

perormance and geometry descriptions. From

that perspective, we posit that heuristics and de-

sign analysis are complements and enact a process

o transorming tacit into explicit knowledge

(Friedman, 2003) o objective perormance phe-

nomena that are later used to generate new design

seeds through additional representations; i these

are then used as active design arteacts, new asso-

ciative leaps and continuous design synthesis can

be achieved.

Figure 21

Principle Perormance Section:

Multivalent Representation,

Iran.


10/10


As a possible consequence or education, design-

ers knowledge o the contributing domains, es-

pecially building science, needs to be improved

by linking it with geometry eects through novel

teaching ormats, as well as research into visual se-

miotics and their relationship to underlying meth-odologies combined with the testing o integrated

design rameworks. A steady accretion o validated

orm-perormance interaces allows the concurrent

expression o engineering and design epistemes;

both need to be acknowledged, regarded in their re-

spective traditions and newly combined to achieve

playul precision. Only then will perormance in-

creases appear easily rom within the design process

itsel.

REFERENCESBatty, WJ and Swann, B 1997, Integration o Computer Based

Modelling and an Inter-Disciplinary Based Approach to

Building Design in Post-Graduate Education, Depart-

ment o Applied Energy, Craneld University, Bedord-

shire, England.

Brown, GZ and DeKay, M 2000, Sun, Wind & Light - Archi-

tectural Design Strategies, 2

nd

. ed., John Wiley & Sons,Hoboken.

Chermaye, S and Alexander, C 1963, Community and Pri-

vacy: Towards a New Architecture o Humanism, Double-

day Anchor Books, New York.

Davis, H 2000, The Culture o Building, Oxord University

Press, Oxord.

Friedman, K 2003, Theory construction in design research:

criteria: approaches, and methods, Design Studies, vol.

24, no. 6, pp. 507-522.

Hetherington, R, Laney, R, Peake, S and Oldham, D 2011, In-

tegrated Building Design, Inormation and Simulation

Modeling: The Need or a New Hierarchy, Proceedings

o Building Simulation 2011, Sydney, Australia, pp. 2241-

2248.

Jakubiec, JA and Reinhart, CF 2011, DIVA 2.0: Integrating

Daylight and Thermal Simulations using Rhinoceros

3D, Daysim and EnergyPlus, Proceedings o Building

Simulation 2011, Sydney, Australia, pp. 2202-2209.

Madsen, M and Osterhaus, W 2005, Exploring Simple As-sessment Methods or Lighting Quality with Archi-

tecture and Design Students, paper presented at the

ANZAScA Conerence, Wellington, New Zealand.

Nasrollahi, F 2009, Climate and Energy Responsive Housing in

Continental Climates, Universittsverlag der TU Berlin,

Berlin.

Palme, M 2011, What Architects want? Between BIM and

Simulation Tools: An Experience Teaching Ecotect, Pro-

ceedings o Building Simulation 2011, Sydney, Australia,

pp. 1410-1430.

Reinhart, CF, Mardaljevic, J and Rogers, Z 2006, Dynamic

Daylight Perormance Metrics or Sustainable Design,

LEUKOS, vol. 3, no. 1, pp. 1-20.

Rittel, HJW and Webber, MM 1973, Dilemmas in a General

Theory o Planning, Policy Sciences, vol. 4, no. 2, pp.

155-169.Smith, A 2004,Architectural Model as Machine: A new view o

models rom antiquity to the present day, Architectural

Press, Oxord.

Strand, RK, Liesen, RJ and Witte, MJ 2004, Resources or

Teaching Building Energy Simulation, Proceedings o

SimBuild 2004, Boulder, USA.

Venancio, R, Pedrini, A, van der Linden, AC, van den Ham, E

and Stous, R 2011, Think Designerly! Using Multiple

Simulation Tools to Solve Architectural Dilemmas, Pro-ceedings o Building Simulation 2011, Sydney, Australia,

pp. 522-529.

building performance modeling in non-simplified architectural design; max c. doelling, farshad...

Documents