10
Building and Environment 37 (2002) 807 – 816 www.elsevier.com/locate/buildenv Design analysis integration: supporting the selection of energy saving building components Pieter de Wilde a ; , Godfried Augenbroe a , Marinus van der Voorden b a College of Architecture, Doctoral Program, Georgia Institute of Technology, Atlanta, GA, USA b Faculty of Architecture, Building Physics Group, Delft University of Technology, Delft, Netherlands Abstract This article motivates the need for more research into the interaction between building design and building analysis in a process context. To provide a context for this discussion, the text focusses on a specic problem: the selection of energy saving building components. A strategy to provide (computational) support for their selection is presented; this strategy is then used to discuss the design support provided by current building analysis tools and to assess probable outcomes of current developments. Finally, a new research project revisiting fundamental issues of design analysis integration is presented. ? 2002 Elsevier Science Ltd. All rights reserved. Keywords: Building design; Performance analysis; Integration; Energy 1. Introduction The interaction between building design and building performance analysis is receiving continuous attention. R&D-projects in the eld are developing new computa- tional methods, (graphical) user interfaces and data models with the IAI-IFC work [1] probably being the most notable ongoing project. But the solutions that are being developed on the interface of design and analysis have yet to prove that interoperability can be achieved in real life projects. A new research project started by Georgia Institute of Technology, Carnegie-Mellon University and University of Michigan, funded by the US Department of Energy, aims to overcome the present limitations of the application of ex- isting performance analysis applications, product modeling techniques, etc., to actual building design. The key to this new approach is a process-centric approach denition of the interfaces. The premises of the project, which is named the Design Analysis Interface (DAI) Initiative [2,3], are: human expertise is an indispensable factor in successful integration of analysis and design; successful execution of data-exchange between analysis tools and data models is impossible without taking the process context into account; Corresponding author. Tel.: +1-404-385-2916; fax: +1-404- 894-1629. E-mail address: [email protected] (P. de Wilde). a process-based approach is needed to enhance the current analysis profession with essential means to better manage and to allow quality control for their performance analysis activities. This relates to capturing audit trails, explicit QA procedures, and workow management. These features will ultimately contribute to the growing maturity of the profession. As building design and building performance analysis cover broad elds, this article focuses on one specic part of the design process (the selection of energy saving build- ing components) and the type of building performance anal- ysis (building energy simulation) that are typically used in their assessment. In doing so the article connects the DAI-Initiative with earlier work at Delft University of Tech- nology in the Netherlands [4,5]. A motivation for focus- ing on the selection energy saving building components and building energy simulation is given below. In contemporary architecture an increasing emphasis on performance aspects like energy consumption and comfort can be observed. The deployment of novel design features like solar walls, advanced glazing systems, sunspaces and photovoltaic arrays is on the rise. Generally the eciency of such energy saving building components cannot be studied in isolation. It is dependent on building character- istics whereas interaction between components can have a substantial eect on the eciency of each individual com- ponent. The impacts of climate conditions and occupant 0360-1323/02/$ - see front matter ? 2002 Elsevier Science Ltd. All rights reserved. PII:S0360-1323(02)00053-7

Design analysis integration: supporting the selection of energy saving building components

Embed Size (px)

Citation preview

  • Building and Environment 37 (2002) 807816www.elsevier.com/locate/buildenv

    Design analysis integration: supporting the selection of energysaving building components

    Pieter de Wildea ; , Godfried Augenbroea, Marinus van der Voordenb

    aCollege of Architecture, Doctoral Program, Georgia Institute of Technology, Atlanta, GA, USAbFaculty of Architecture, Building Physics Group, Delft University of Technology, Delft, Netherlands

    Abstract

    This article motivates the need for more research into the interaction between building design and building analysis in a process context.To provide a context for this discussion, the text focusses on a speci/c problem: the selection of energy saving building components.A strategy to provide (computational) support for their selection is presented; this strategy is then used to discuss the design supportprovided by current building analysis tools and to assess probable outcomes of current developments. Finally, a new research projectrevisiting fundamental issues of design analysis integration is presented. ? 2002 Elsevier Science Ltd. All rights reserved.

    Keywords: Building design; Performance analysis; Integration; Energy

    1. Introduction

    The interaction between building design and buildingperformance analysis is receiving continuous attention.R&D-projects in the /eld are developing new computa-tional methods, (graphical) user interfaces and data modelswith the IAI-IFC work [1] probably being the most notableongoing project. But the solutions that are being developedon the interface of design and analysis have yet to provethat interoperability can be achieved in real life projects.A new research project started by Georgia Institute of

    Technology, Carnegie-Mellon University and University ofMichigan, funded by the US Department of Energy, aims toovercome the present limitations of the application of ex-isting performance analysis applications, product modelingtechniques, etc., to actual building design. The key to thisnew approach is a process-centric approach de/nition of theinterfaces. The premises of the project, which is named theDesign Analysis Interface (DAI) Initiative [2,3], are:

    human expertise is an indispensable factor in successfulintegration of analysis and design;

    successful execution of data-exchange between analysistools and data models is impossible without taking theprocess context into account;

    Corresponding author. Tel.: +1-404-385-2916; fax: +1-404-894-1629.

    E-mail address: [email protected] (P. de Wilde).

    a process-based approach is needed to enhance the currentanalysis profession with essential means to better manageand to allow quality control for their performance analysisactivities. This relates to capturing audit trails, explicit QAprocedures, and workEow management. These featureswill ultimately contribute to the growing maturity of theprofession.

    As building design and building performance analysiscover broad /elds, this article focuses on one speci/c partof the design process (the selection of energy saving build-ing components) and the type of building performance anal-ysis (building energy simulation) that are typically usedin their assessment. In doing so the article connects theDAI-Initiative with earlier work at Delft University of Tech-nology in the Netherlands [4,5]. A motivation for focus-ing on the selection energy saving building components andbuilding energy simulation is given below.In contemporary architecture an increasing emphasis on

    performance aspects like energy consumption and comfortcan be observed. The deployment of novel design featureslike solar walls, advanced glazing systems, sunspaces andphotovoltaic arrays is on the rise. Generally the eHciencyof such energy saving building components cannot bestudied in isolation. It is dependent on building character-istics whereas interaction between components can have asubstantial eIect on the eHciency of each individual com-ponent. The impacts of climate conditions and occupant

    0360-1323/02/$ - see front matter ? 2002 Elsevier Science Ltd. All rights reserved.PII: S0360 -1323(02)00053 -7

  • 808 P. de Wilde et al. / Building and Environment 37 (2002) 807816

    behavior add to the complexity and make it almost impos-sible to predict performance without use of computationaltools.The building physics profession has a large number of

    energy-related computational tools at its disposal, rangingfrom simple to sophisticated computer programs, and theaspects that are considered may range from one aspect onlymore integral views. Most tools allow comparison of variousdesign options under identical conditions.Earlier research [4] dealt with the integration of energy

    saving building components in real-life building design sce-narios as well as the use of computational tools in these sce-narios. The results indicate that most energy saving build-ing components are selected without proper underpinning,and that computational tools only play a limited role in theselection of these components, mainly due to lacking rigorand adequate information during the average design deci-sion process. The crucial design phase for the selection ofenergy saving building components appears to be the phaseof conceptual design. The need expressed in [4] is that fu-ture design tools should be able to respond to procedures(process templates) for speci/c tasks of the building de-sign process. Appropriate computational tools (both existingand future tools) should be associated with these speci/ctasks [4,5].

    2. Approach

    The main goal of this article is to provide a set of argu-ments for the importance of treating the interaction of build-ing design and building performance analysis in a processcontext. The objective is to show that current eIorts on theinterface between design and analysis alone are not goingto achieve interoperability.The second objective is to address the design problem of

    selection of energy saving components in practice. The ob-jective is to develop a strategy that supports rational deci-sions with respect to the integration of energy saving com-ponents in building design. Such a strategy must addressthe aforementioned issues and direct and manage the useof computational tools for the analysis tasks. The strategyshould especially be applicable during the conceptual de-sign phase. It is acknowledged however that the selectionof energy saving technologies cannot be represented as apurely rational design decision making process. The limita-tions of this view have been demonstrated in architecturaldesign studies where it has been shown that the design pro-cess is at times highly intuitive and unpredictable and there-fore hardly rational. It can be expected that potential con-Eicts and group dynamics of the design team will interferewith rational choices of technologies for optimal buildingperformance. In our research, these issues are not consid-ered, as the focus is on providing objecti/ed performanceresults at the right time to the design team in order to betterinform the design decision making process.

    This article takes a bottom-up approach. It starts by devel-oping the strategy for the selection of energy saving build-ing components based on the existing body of knowledge;the strategy describes how selection of energy saving com-ponents should take place in theory. Using this strategy thedesign support provided by current building performanceanalysis tools is assessed, as well as the improvements thatcan be expected from the solutions resulting from ongo-ing R&D eIorts on new computational methods, (graphical)user interfaces and data models. Then the article will take abroader scope and present the approach of the DAI-Initiativein developing a process-centric prototype supporting the in-teraction of design and analysis in general.The following paragraph discusses development of the

    strategy for selection of energy saving building compo-nents. This strategy is based on the concepts of systemsengineering as described by Hazelrigg [6] and Blan-chard and Fabrycky [7], augmented with theories fromengineering design as presented by Cross [8] and decisiontheory, described by Keeney and RaiIa [9]. The under-lying fundamentals from these disciplines will be brieEydiscussed. Then these concepts will be applied to that partof the building design process that deals with the selectionof energy saving building components (with emphasis onthe conceptual design phase).The next paragraph discusses support for selection of

    energy saving components. Speci/c needs for design andanalysis support are de/ned. These needs are then con-fronted with currently existing building energy analysistools and solutions expected from ongoing R&D eIorts.Needs not covered by existing tools or solutions under de-velopment will be apparent targets of the ongoing researchproject.The /nal paragraph discusses the DAI-Initiative. Its back-

    ground, major objectives and approach will be presented.

    3. A strategy for the selection of energy saving buildingcomponents

    3.1. Underlying fundamentals

    3.1.1. Systems engineeringSystems engineering can be de/ned as the application

    of scienti/c principles to the design, development, imple-mentation and control of systems, where a system is a setof interrelated components working together towards somecommon objective or purpose. System engineering can beapplied to many disciplines; however, though there is gen-eral agreement on the principles and objectives of systemengineering, each application will strongly depend on thisdiscipline and the background and experiences of the partic-ipants, as well on the complexity of the system [7,10]. Theapplicability of systems engineering to building design hasbeen demonstrated in the context of cost control by Merritand Ambrose [11].

  • P. de Wilde et al. / Building and Environment 37 (2002) 807816 809

    Systems engineering addresses all major life-cycle pro-cesses of systems: systems design, development, pro-duction=construction, distribution, operation, maintenance=support, phase-out and disposal; in this article the emphasiswithin systems engineering will be on (conceptual) buildingdesign.According to the systems engineering approach the de-

    sign process encompasses four distinct activities: the anal-ysis of objectives and constraints; the identi/cation of alldesign options that are to be considered; the developmentof expectations on outcomes for each option; and the use ofvalues to select the option that has the range of outcomesand associated probabilities that are most desired.The /rst activity consists of determination of the objec-

    tives that the design team wants to achieve and the determi-nation of limiting factors (constraints). For each objectiveand=or constraint attributes must be identi/ed that describethe extent to which these objectives are being achieved orconstraints are being met. Attributes are also known as (tech-nical) performance measures or more generally as perfor-mance indicators; in this article the last term will be used. Itis important that the list of performance indicators is com-plete (adequately covers the objective), operational (mean-ingful) and non-redundant (preventing double counting ofachievements). Additionally the set of performance indica-tors should help to decompose the objective(s) into manage-able parts, while at the same time being of minimum size.Limiting values for speci/c performance indicators are

    called goals or requirements. Goals are values that the de-sign team strives to achieve; requirements are values that adesign must meet in order to be acceptable. Both goals andrequirements are either ful/lled or not.Note that in most design projects objectives, constraints,

    attributes, requirements and goals are not present from theoutset. Normally the client provides only general needs andwishes that must be translated into objectives, requirementsand goals. Constraints are frequently dependent of actualdesign proposals and hence need to be de/ned during thecourse of the design process.The second activity is identi/cation of all design options

    that are to be considered; in systems engineering the set ofall these options is named the option space. De/nition ofan option space requires design synthesis or the de/nition ofsystem con/gurations, and parameterization or identi/cationof the parameters and their permissible ranges.An option space can be constructed by eliminating possi-

    bilities from the set of all possible constructs, or by startingfrom one construct and adding options for consideration.Adding options is an appropriate approach when using ex-isting components to design new systems [6].The third activity is the development of expectations

    on outcomes for each design option. Outcomes are theachievements of the individual options for objectives andconstraints, described using the performance indicators.Theoretically the outcomes should be expressed in termsof predicted achievements and associated probability of

    occurrence, because of uncertainty in the design itself, theconditions in which the design will perform, and the pre-diction method. Although this is not always feasible, recentprogress in assessing uncertainties in building performanceassessments is reported by de Wit and Augenbroe [12]. Theset of outcomes is named outcome space.The fourth and /nal activity is selection of the design op-

    tion that has the range of outcomes (and associated probabil-ities) that are most desired. This involves determination ofthe values of all design options. A value indicates the attrac-tiveness of each design option in relation to the objectivesand constraints. If this value is purely based on objecti/edperformance measures, it is called utility. If overlaid bythe preferences or value system of the observer it is moreappropriate to call it quality.Based on the obtained values one design option must be

    chosen. This involves mostly a decision problem with sev-eral objectives and=or constraints, determination of prefer-ences and value trade-oI. It can be advisable to use a formaldecision method (weighting or utility function) that supportsthe decision maker by dividing one complex choice probleminto several simpler choice problems [9,13].The four activities will reoccur many times during one

    design process. New objectives and constraints will emerge,new options will be added and evaluated, and the designwill develop by a chain of design decisions.

    3.1.2. Engineering designSystem engineering as a whole provides a too general ap-

    proach. For application to the technical design of buildings,speci/c methods need to be selected and tailored.One engineering design method that will be used is the

    objective tree method which shows the design objectivesin a diagrammatic form; it clari/es the relationships be-tween objectives as well as the hierarchy of objectives andsub-objectives. The approach for developing an objectivetree consists of /rst identifying areas of concern; withinthese areas of concern, objectives can be identi/ed, and ifneeded lower-level objectives can be identi/ed within theobjectives.Principles from engineering design will also be used to

    support the step of design synthesis. This step consists ofthe de/nition of system con/gurations or, in other words,generation of alternatives; it is the creative part of the pro-cedure. However, engineering design theory states that cre-ativity often consists of reordering or recombination of ex-isting elements into a wide range of variants, often resultingin novel solutions. The use of existing elements has the ad-ditional advantage that it allows the design team to avoidsome detailed design work. It also eliminates uncertaintyfrom an area of the overall design, as the speci/cation ofthe element in question is known. However, the use of ex-isting elements leaves aside the creation of novel elements.A method that helps to generate alternative design solutionsfrom known elements is the morphological chart method;

  • 810 P. de Wilde et al. / Building and Environment 37 (2002) 807816

    it consists of listing essential functions of the design un-der development, listing means (the existing elements) bywhich these functions might be achieved, and combiningthese means to de/ne the total search space (the set of allpossible option spaces) for this design project [8].

    3.1.3. Decision theoryDecision theory is concerned with making rational choices

    between alternatives by applying (mathematical) methods.Within decision theory a distinction is made between single-and multiple-attribute decision problems, depending on thenumber of descriptors (attributes) that are needed to spec-ify the consequences of a decision. Also a distinction ismade between problems under certainty or uncertainty. Forproblems under certainty the consequence(s) of a decisionare known; for problems under uncertainty there is a rangeof possible consequences. Making decisions under uncer-tainty involves taking (or limiting) risks. Decision methodsthat provide a preference order are named ordinal methods;methods that provide a preference order as well as a mea-sure of the strength of these preferences are named cardinalmethods [9].The most widely used method suitable for multiple-

    attribute decisions under certainty consists of an additiveutility function and is described by

    U (Ai) =m

    j=1

    ieij;

    where U (Ai) is the utility of alternative Ai with regard to allcriteria C1; : : : ; Cm, Ai the alternative i, i = 1; : : : ; n, Cj thecriterion j, j = 1; : : : ; m, j the weighting factor of criterionCj, representing the importance of Cj to the overall utility,and eij the eIectiveness of alternative Ai related to criterionCj (after Ref. [13]).

    3.2. A framework for the selection of energy savingbuilding components

    Following the systems engineering approach the proce-dure for the selection of energy saving building componentsconsists of the following four main activities:

    1. analysis of objectives and constraints that control the se-lection of components, and speci/cation of appropriateperformance indicators;

    2. development of an option space that consists of com-binations of building design(s) and energy saving build-ing components and a parameterization of these combi-nations;

    3. determination of the performance of these combinations;4. selection of the most desirable combination of building

    design and components.

    3.2.1. Analysis of objectives and constraintsIn current practice the starting point for the selection of

    energy saving components is found to be underdeveloped.

    good building design

    comfort

    functionality

    safety

    architectural value

    environmental impact

    financing

    thermal comfort

    visualcomfort

    acoustical comfort

    air quality

    vibration control

    emmissions

    water use

    energy use

    material use

    Fig. 1. Objective tree for building design (reduced).

    In most cases, the design brief contains only few instructionsconcerning energy use (which is appropriate as the brief isjust a short description of the kind of building design that isto be developed); these instructions might vary from a state-ment of a broad overall objective (develop an environmen-tally friendly building) to reference to building codes orspeci/cation of clear requirements concerning energy use.However, design teams do not seem to take the time to de-velop these instructions into a proper list of objectives andconstraints, performance indicators, and requirements andgoals. The lack of this list results in heuristic, single-attributedecision making on energy saving components [4].Obviously, the main objective of the use of energy

    saving building components is to make buildings moreenergy-eHcient. Yet energy eHciency is only one ofmany objectives that must be considered in building de-sign; the notion of energy eHcient building design as amono-discipline is clearly /ctitious. The objectives that arerelevant when making design decisions concerning energysaving building components can be mapped using a (red-uced) general objective tree for building design; see Fig. 1.For energy saving building components the main objec-

    tive is eHcient use of energy; all other building design objec-tives act as constraints. Of these constraints thermal comfortis most related to energy use. Both energy use and thermalcomfort are highlighted in the objective tree.To capture these objectives and constraints, an extensive

    set of performance indicators can be used. An overview ofsome of the possible performance indicators for the objec-tive energy eHciency and the constraint thermal comfortis presented in Table 1. Note that these performance in-dicators are based on an evaluation of the design under agiven experiment. Performance indicators are functions ofdesign properties (e.g. U -value), whereas U -value is not a

  • P. de Wilde et al. / Building and Environment 37 (2002) 807816 811

    Table 1Overview of some performance indicators for energy eHciency andthermal comfort

    Objective Perf. indicator Symbol Unit

    Energy Energy consumption E J or GJeHciency Peak heat demand q J=s

    Thermal Predicted mean vote PMV comfort Predicted percentage PPD %

    of dissatis/edDiscomfort degree hours ddh hMinimum or Tmin

    C or K

    maximum temperature Tmax

    performance indicator by itself. Also note that further spec-i/cation of these performance indicators is possible (for in-stance one can discern energy consumption for whole build-ings or for rooms, and energy consumption for heating andfor cooling). See Table 1.

    3.2.2. Development of an option spaceStudy of recent design projects reveals that in most

    projects an option space for energy saving building com-ponents is virtually non-existent: 8090% of all compo-nents are selected without consideration of an alternative[5]. Hence, design teams should be stimulated to broadentheir search. An obvious strategy is to generate alternativesystem con/gurations by combining a given building designwith one or more components.In such a strategy for the selection of components it is

    necessary to consider the building design to be invariant andleave aside the earlier development of this design, hence treatthe design information as an input to the procedure. Even asinput only the building design can take many shapes, froma design that is de/ned by a volume and a building functiononly up to a completely de/ned existing building for whichall details are known (as encountered in renovation projects).However, as the most important phase for the selection ofcomponents is early conceptual design, this is the designdevelopment stage that lies at the heart of this study.Building components are existing technical solutions;

    building designers can choose from an array of possibili-ties. Yet, a random search for an energy saving componentmostly leads to consideration of many inappropriate com-ponents; a morphological chart can help to arrive quickly ata number of components that might be useful for a speci/cbuilding design. A total of nine main functions of energysaving components can be discerned: limitation of transmis-sion losses, limitation of ventilation and in/ltration losses,energy storage, utilization of internal heat loads, use of re-newable energy (solar, wind, hydro, etc.), eHcient heating,eHcient cooling and miscellaneous (like inEuencing occu-pant behavior). Approximately 90 energy saving compo-nents are available to achieve these functions. As exam-ple, the storage function part of the morphological chart isshown in Fig. 2.

    storagewall

    earthstorage

    aquiferthermalmass

    waterreservoir

    means / energy saving component

    function: energy storage

    chemicalstorage

    phasechange

    remotestorage wall

    Fig. 2. Morphological chart for energy saving components showing thefunction energy storage.

    Using the morphological chart the design team can makea pre-selection of energy saving building components. Thispre-selection remains based on (subjective) preferences;without further evaluation of the performance of the com-ponents no well-founded /nal selection is possible. Notehowever that evaluation of all possible combinations is notwithin reach: even for a building with four energy savingcomponents (the average number) the 90 components allowto make more than 60 000 000 combinations [5].For each type of energy saving component a set of rele-

    vant parameters can be de/ned. The range of these param-eters is either de/ned by the building design or by externalfactors; for instance the maximum area of the separatingwall between a sunspace and the adjacent building is limitedby the building design, whereas the COP (coeHcient of per-formance) of a heat pump depends on physics. Parametersthat relate to the building design are named design depen-dent parameters; the other parameters are named designindependent parameters.A catalogue of energy saving components that gives the

    design team easy access to design-dependent parameters,design-independent parameters and their limiting factorswill help to speed up the de/nition of the option space.

    3.2.3. Determination of the performance of all optionsDetermination of the performance of a building is the tra-

    ditional /eld of building performance simulation; however,application of building simulation tools to real-life buildingdesign remains problematic. In actual design projects sim-ulation tools are mainly used for optimization and veri/ca-tion, and not to support basic design decisions. Furthermore,the selection of tools seems to be arbitrary; there is no sup-port system in place to guarantee that an appropriate tool isused for a speci/c design evaluation [5]; yet, one tool onlymust be picked from a whole range like presented on theDOE energy tool directory on the internet [14].Application of computational tools to support the selec-

    tion of components should be aligned with the objectivesand constraints that have been identi/ed, and within the op-tion space that has been developed. It should also anticipatethe selection process that will follow. This can be realized by

  • 812 P. de Wilde et al. / Building and Environment 37 (2002) 807816

    atrium

    climate facade

    double facade

    sunspace

    Eann PPDPMV $

    opt

    ion

    spac

    eperformance

    indicators

    Fig. 3. Example of a performance matrix.

    Fig. 4. Use of an enlarged performance matrix to support the selectionof an appropriate (set of) tool(s).

    setting up a performance matrix, which can be transformedinto an assessment matrix during the next step. The columnsof the performance matrix are de/ned by the performanceindicators that capture the objectives and constraints; therows are de/ned by the elements of the option space (theenergy saving components). Now a tool (or combination oftools) that is able to /ll this matrix must be selected. SeeFig. 3.To support the search for tools an enlarged performance

    matrix can be used that includes all available performanceindicators and all available options (the whole search spaceof all energy saving components). In this enlarged matrixit is possible to map all cells that can be /lled by individ-ual simulation tools; the result is a footprint of these tools.Now the cells de/ned by a set of selected performance in-dicators and an option space can be used as a /lter to /ndan appropriate (set of) tool(s). See Fig. 4.The accuracy of the output produced by the tools must be

    suHcient to discern the requirements and goals as de/ned. Asurplus of accuracy is easily discarded during the transfor-mation of the performance matrix into the assessment ma-trix; however, this implies that unnecessary computationaleIorts have been made.

    It is important to note that the output of tools is a perfor-mance prediction based on a set of assumptions (concerningbuilding design=operations=climate, etc.), modeling eIortsand computational operations; all these introduce their ownuncertainties. However, when it comes to a risk and uncer-tainty assessment of building performance assessments, wehave to acknowledge that this /eld has been receiving at-tention only recently; for more information, see Ref. [15].As uncertainty analysis for building simulation still needsto go mainstream, it is not yet realistic to add uncertainty tothe framework.

    3.2.4. Selection of the most desirable optionIn current practice, the selection of energy saving com-

    ponents is found to be highly intuitive; the choice for a spe-ci/c component is mostly based on experience and analogy.However, intuition and experience are neither absolute norobjective; moreover, good intuition and experience cannotbe transmitted to succeeding generations [5].The /nal step in the discussed framework involves the

    evaluation of all data in the performance matrix and a se-lection of the most desirable design option (energy savingbuilding component) based on this evaluation; this makesthe selection procedure more rational, transparent and opento discussion.First, all options must be checked for meeting all require-

    ments; options that do not meet a requirement are ruled out.A preference for one of the remaining options must be basedon the extent to which these options meet the goals. This im-plies that a (subjective) value must be assigned to the data inthe performance matrix, thus transforming it into an assess-ment matrix. In the assessment matrix the extent to whichthe options meet the diIerent goals must be measured onthe same scale, allowing to compare diIerent performanceaspects. For instance energy consumption, thermal comfortand costs can all be measured according to the scale:

    0:0 = does not meet the goal;0:5 = just meets the goal;1:0 = perfectly meets the goal.

    Selection of the most desirable option from the result-ing assessment matrix involves making a tradeoI betweenthe relevant performance aspects. This tradeoI is subjectiveonce more. However, by using an additive utility functionthe underlying subjective value structure is made explicitand negotiable. A /ctitious example of an assessment matrixwith weighting factors and overall utility values is shown inFig. 5. The option with the highest value will be selected.

    4. Support for the selection of energy saving buildingcomponents

    4.1. Speci9c needs for support of the selection of energysaving building components

    If energy saving building components are to be se-lected according to the strategy described in the previous

  • P. de Wilde et al. / Building and Environment 37 (2002) 807816 813

    climate facadedouble facade

    sunspace

    Eann K ($)

    opt

    ion

    spac

    e

    performance indicators

    atrium

    PMV U(A)50 20 30 (100)

    1.0 1.0 0.0 701.00.50.5

    1.00.50.0

    0.51.01.0

    856555

    weighting factor

    Fig. 5. Use of an additive utility function to select the most desirableoption from an assessment matrix.

    paragraph, the following elements would need to be takencare of in a design decision support system:

    objective tree, relevant for energy saving building com-ponents;

    overview of all performance indicators per objective; morphological chart with all available energy savingbuilding components;

    overview of relevant design-dependent and design-independent parameters of each component;

    support for the selection of building analysis tools, usingfootprints of existing tools;

    decision rules, e.g. additive utility theory.

    Any support system for the selection of energy saving com-ponents must be capable of following the highly Eexible anditerative design process; also it should be able to supportdiIerent users (project-speci/c design teams).Building performance analysis tools supporting the strat-

    egy for the selection of energy saving components (whetherstand-alone or embedded in a support system) must be ableto predict the performance of design options that consist ofthe combination of a given building design and one or moreenergy saving components. This results in the following re-quirements:

    Analysis tools are to accept a variety of building designsas inputs, and especially be able to deal with conceptualdesigns. Basically, this means that these tools must beable to capture the essentials of a design whether thisdesign is only represented by a few major properties oris described in detail.

    The framework assumes that the option space is de/nedby a pre-selection of existing energy saving building com-ponents. Evaluation of this option space is facilitated byusing tools that have pre-de/ned models of energy savingcomponents which can easily be invoked. Energy savingcomponents can be described in outlines or in detail, too.Note that sophisticated computational tools can be usedwith both coarse and re/ned models, whereas simple toolsintrinsically only cover coarse or lumped models.

    The output of computational tools must /t in the perfor-mance matrix; moreover, it must have an accuracy that al-lows one to discern requirements and goals. This has thefollowing consequences:

    Tools should allow one to select and=or modify perfor-mance indicators so that they correspond with actual ob-jectives and constraints.

    The performance data generated by tools must allow eval-uation and post processing in order to incorporate thisdata into an overall value assessment.

    Adjustable accuracy can be useful to bring the computa-tional eIort in line with required results.

    4.2. Support provided by current building energy analysistools

    The current generation of energy analysis tools is notconcerned with supporting design strategies. Only veryfew tools oIer some facilities that can be viewed as su-per/cial /rst developments towards design decision sup-port. A good example are the functions provided by theEnergy-10 program [16]. Energy-10 oIers some supportfor the development of an option space through a set ofenergy saving strategies that can be applied to a givendesign. However, it does not include a full morphologicalchart, while energy saving components come in sets inthe strategies, making it diHcult to evaluate single compo-nents. Energy-10 also provides an order of energy savingstrategies based on achieved energy savings; however,this rank-feature does not allow multiple design objec-tives, as it only has a one-criteria decision rule. In orderto provide the above functions, Energy-10 severely limitsthe possibilities to describe a given design: all buildingsare de/ned as a box; only a limited number of buildingfunctions within a speci/ed range of building sizes can beevaluated.As for using existing building energy analysis tools within

    the strategy for selection of energy saving components, thefollowing observations can be made:

    The current set of tools covers a broad range of buildingdesigns. Given enough time and resources almost anydesign can be analyzed. However, only expert knowledgecan guarantee that essential features of a given design areactually included in the analysis.

    At the moment there is no tool that has a footprint thatcovers the whole search space (all energy saving buildingcomponents) as well as all possible performance indica-tors; neither is development of such a tool to be expectedwithin the next two or three decades. This means thatin most cases a combination of computational tools willhave to be used.

    Alternative design options are to be compared on the basisof equivalence. Consequently the use of one tool to de-termine the performance for one aspect (performance in-

  • 814 P. de Wilde et al. / Building and Environment 37 (2002) 807816

    dicator, one column in the performance matrix) is prefer-able; the use of more than one tool is only possible whentaking into account the diIerences due to (physical) mod-eling and computational procedures. In the /eld of energysimulation and related aspects only a few tools cover asubstantial part of the whole search space for energy sav-ing building components: TRNSYS [17], ESP-r [18] andEnergyPlus (DOE-2=BLAST) [19] all allow evaluation ofa fairly broad range of energy saving components (avail-able in these tools as pre-de/ned models), which mightexplain part of the success of these tools.

    4.3. Support to be expected from the solutions resultedfrom ongoing R&D e=orts

    Throughout this paper three main ongoing R&D eIortson the interface of design and analysis have been identi/ed:development of new computational models, (graphical) userinterfaces and data=product models.

    New computational models will have only a limited im-pact on the role of tools in building design. They willonly add to the broad range of tools and models alreadyavailable. New models that claim to be design tools be-cause they work with a limited set of input parameterswill only be applicable to design problems where exactlythis set of parameters is relevant (a solution looking fora suitable problem). New detailed computational mod-els can allow better analysis of speci/c components orbehavioral aspects, but such models do not close the gapbetween analysis and design.

    Development of (graphical) user interfaces is a good wayto makes analysis tools easier to use. However, the im-pact on using tools for building design is limited. Staticinterfaces that hide the complexity of tools through theuse of default values constrain the use of tools in the sameway that simpli/ed computational models do. We arguethat only process-sensitive, case-speci/c user driven in-terfaces can be expected increase the support of real-lifebuilding design analysis teams dramatically.

    Data=product models that allow automation of informa-tion exchange between design tools (CAD) and analy-sis tools have been getting a lot of attention since theearly 1980s [20]. If fully automated information exchangecould be achieved between analysis and design applica-tions, it would have a huge impact on design support.However, two factors obstruct this approach: (1) it is dif-/cult to develop a data model that includes all buildingdesign information; there will always be a large percent-age of unstructured data that is not included in currentinformation models and needs to be transferred throughintuition, interpretation and expert judgment; and (2) asanalysis tools in diIerent domains may have very diIerentinformation requirements; it seems doubtful that informa-tion can be mapped from one domain to the other. More-over, it only makes sense to evaluate the performance of

    the design options for various performance aspects (en-ergy, lighting, acoustics, costs, etc.) if and only if theseperformance aspects are related to actual design objec-tives and constraints expressed by the design team at thattime. Hence, the usefulness of product models that facili-tate data exchange between diIerent computational toolsis strongly dependent on the context of design decisionsto be made. Clearly automation of this data transfer alonewill not be the primary enabler of computational tools indesign; moreover, the data transfer between tools mightbe too complicated to solve in general. In this light thedevelopment of integrated design environments should beguided by principles of openness, Eexibility, user drivenmappings and data selection, version control and consis-tency management, and above all allow (remote) multi-actor collaboration.

    5. Providing support for the interaction of design andanalysis: the Design Analysis Interface (DAI) Initiative

    5.1. Background

    The DAI research initiative responds to the need todevelop credible solutions for deploying and integratingenergy analysis tools in the building design process. Theunderlying objective is to enable a more eIective and eH-cient use of existing and emerging analysis tools by buildingdesign and engineering teams. The fundamental goal is tofacilitate the work of design professionals in their use ofanalysis tools. Building modeling, integration and standard-ization have been thought to be strong facilitating technolo-gies. Based on evaluation of projects dealing with thesetechnologies, the authors contend that available solutionsfall far short of the promises associated with the availabilityof integration technologies in general. The circumstances,that render this initiative urgent and timely are reviewed inmore depth in an earlier white paper [3], dealing with ad-vances in Product Data Technology (PDT) in general andits adequacy for design analysis integration in particular.

    5.2. Goal

    The goal of the DAI-Initiative is to provide a layer of sup-port between an analysis request and a particular simulationtool. The scenario layer enables the consultant to concen-trate on the analysis by selecting analysis functions from apalette of generic tool functions, without worrying whichparticular software product provides that functionality. Thisis a logical step towards tool independence and easy adap-tation to future simulation environments.The outcome of the research will additionally provide

    useful functions:

    storing an audit trail of the analysis; easy reuse in future projects; explicit Q.A. procedures;

  • P. de Wilde et al. / Building and Environment 37 (2002) 807816 815

    workEow management; training tool for novices.

    The DAI project targets a prototype that satis/es the fol-lowing requirements:

    A user-centric approach: The approach to the systemshould be user-centric and thus reEect the new realityin integrated systems, which dictates that no eIective in-tegration can exist without the expert user in the loop.Rather than static data interfaces between design and anal-ysis representations, the interface will provide a work-bench to the energy consultant that supports the uniqueskills of transforming design information, evaluation re-quest, experience, tool limitations, budget constraints etc.into the most adequate simulation model.

    A work?ow centric approach to interaction events be-tween design and analysis. Each design analysis interac-tion is speci/cally tied to a purpose and a need toonthe one hand acquire speci/c aspects of the design, it in-tentions and contextand on the other to communicateperformance characteristics of the design at that stage.This requirement stems from the observation that no onesize /ts all data interface can be de/ned that would takecare of all the parameters that de/ne the interaction con-text. Each design analysis interaction will normally in-volve multiple atomic evaluations, the sequence of whichwill be explicitly de/ned in a scenario. Without strongknowledge of these scenarios, the goals of data exchangeremain unde/ned.

    Explicit de9nition and management of evaluationscenarios. A scenario is a task Eow model of a speci/cevaluation or condition. A scenario model de/nes the re-lationships and Eow logic between the tasks that are partof the scenario. Tasks are either atomic evaluation tasks(user or application performed) or supporting tasks,like checking procedures, post processing of analysisdata, design interrogation tasks, etc. The Eows are notsimple sequences but complex sequences of states, withbranches and iteration and possibly recursion. A partic-ular scenario model de facto represents the intra-oHcework Eow execution of the requested evaluation. It makeslocal interoperability requirements explicit.

    Scenario speci9c building simulation model interfaces,involving the semantic uni/cation of all information Eowsinternal to one or more scenarios. A building simulationmodel and its interfaces should oIer the interoperabilitysupport for scenario tasks and user interactions. It shouldbe recognized that diIerent building simulation modelsmay exhibit considerable overlap that might lead to futuremore generic building simulation models.

    Easy generation of internal data interfaces, i.e. be-tween the simulation model and the tasks within a sce-nario. Current interface toolkits are too complicated andtoo slow for production use. A new interface develop-ment kit is needed that allows data interfaces to be Eexi-

    bly de/ned and easily generated. They should be able torespond to user supplied conditions regarding simulationtool use and the work Eow between tools.

    A tool independent system architecture. Interfaces shouldbe built to analysis functions, rather than to software tools.It should be easy to substitute new tools when they arrive(most notably emerging OO simulation environments) orto add new tools. Here, interfaces refers to the user in-terface, the system and data exchange interfaces. This re-quires that atomic evaluation functionality, data formats,and user interfaces all should be loosely coupled with theapplications that perform them.

    Support of incremental design analysis cycles impliesthat the scenarios should be able to de/ne and supportdesigner interaction tasks facilitating both one-on-one de-signer consultant interaction and iterated cycles of explo-ration. These involve repeated simulation cycles with in-cremental changes to conditions or to the design. Resultsneed to be easily recorded and tracked. These types of in-teraction should enable parameter selections and view /l-ters on the performance data in building simulation mod-els, supporting hill-climbing and other forms of perfor-mance improvement.

    User controlled gateways to design information shouldprovide access to and easy integration of both traditionalCAD geometry and data and also unstructured designinformation from product libraries, catalogs and otherdatabases. It should facilitate using this diverse informa-tion to populate a building simulation model with the ap-propriate data. These gateways should be user controlledand not semantically limited to the /xed semantics of anexisting design building model. The gateways should of-fer a workbench-like environment that provides access toa variety of design information sources in heterogeneousformats (STEP, CSI, DXF, IAI, P-LIB, Sweets Catalog,ASHRAE data, VRML, etc.), oIering a set of functionsto select, extract, map and structure meaningful informa-tion from these sources to the instances in the buildingsimulation models. These so-called entry and exit gate-ways should support and enhance, not replace the vitalexpert skills and tacit knowledge of the consultant whenperforming analysis modeling and schematization. Thegateways should take advantage of the recent progress inexpandable middle ware access layers to external het-erogeneous design information for human inspection, se-lection and post processing. This facilitates a workbenchapproach that we believe should become the next gener-ation of interfaces between design building models andsimulation building models.

    The /rst year of the DAI-Initiative consists of the de-velopment of a mock-up of the intended product, whichdemonstrates the major functions of the envisioned de-sign analysis interface. At the end of the /rst phase theDAI-products will be demonstrated and tested in work-shops with the user-community. Actual results from the

  • 816 P. de Wilde et al. / Building and Environment 37 (2002) 807816

    DAI-Initiative are posted on the Internet, where productsand material for further reading can be downloaded [2].

    6. Conclusion

    This article motivates the need for more research into theinteraction between building design and building analysis ina process context by discussing a speci/c design decisionproblem: the selection of energy saving building compo-nents. A strategy to provide adequate computational supportfor this design decision is presented; this strategy demon-strates how concepts from systems engineering, engineeringdesign and decision theory can be used to develop a Eexibleprocess support for the deployment of computational toolsto inform rational support design decisions. Possibilities ofcurrent building energy analysis tools as well as solutionsexpected from ongoing R&D-eIorts to support this strat-egy are reviewed in order to identify areas that need furtherattention.The DAI Initiative is presented as research project that

    responds to the need to develop credible solutions for de-ploying and integrating energy analysis tools in the buildingdesign process. The underlying objective is to enable a moreeIective and eHcient use of existing and emerging analysistools by building design and engineering teams. The funda-mental goal is to facilitate the work of design professionalsin their use of analysis tools. The discussion of the DAI re-search project makes clear that solutions to support the se-lection of energy saving building components will need amajor eIort and are not available yet.

    References

    [1] International Alliance for Interoperability website. URL:http://www.iai-na.org/

    [2] Design Analysis Interface Initiative website. URL: http://dcom.arch.gatech.edu/dai/

    [3] Augenbroe G, Eastman C. Needed progress in building designproduct models. White paper, 2000.

    [4] de Wilde P, Augenbroe G, van der Voorden M. Invocation ofbuilding simulation tools in building design practice. Proceedings ofthe Building Simulation 99, Sixth International IBPSA Conference,Kyoto, Japan, September 1315, 1999. p. 12118.

    [5] de Wilde P, van der Voorden M, Brouwer J, Augenbroe G, Kaan H.Assessment of the need for computational support in energy-eHcientdesign projects in the Netherlands. In: Proceedings of the BuildingSimulation 2001, Seventh International IBPSA Conference, Rio deJaneiro, Brazil, August 1315, 2001.

    [6] Hazelrigg GA. Systems engineering: an approach toinformation-based design. Upper Saddle River: Prentice-Hall, 1996.

    [7] Blanchard BS, Fabrycky WJ. Systems engineering and analysis.Upper Saddle River: Prentice-Hall, 1998.

    [8] Cross N. Engineering design methodsstrategies for product design.Chichester: Wiley, 2000.

    [9] Keeney RL, RaiIa H. Decisions with multiple objectivespreferences and value tradeoIs. Cambridge: Cambridge UniversityPress, 1993.

    [10] INCOSE website, International Council On Systems Engineering.URL: http://www.incose.org/.

    [11] Merrit FS, Ambrose J. Building engineering and systems design.New York: Van Nostrand Reinhold, 1990.

    [12] de Wit S, Augenbroe G. Uncertainty in building design analysis. In:Proceedings of the Building Simulation 2001, Seventh InternationalIBPSA Conference, Rio de Janeiro, Brazil, August 1315, 2001.

    [13] Roozenburg N, Eekels J. Productontwerpen, structuur en methoden.Lemma, Utrecht, 1991 (in Dutch).

    [14] DOE website, US Department of Energy Building EnergySoftware Tools Directory. URL:http://www.eren.doe.gov/buildings/tools directory/

    [15] De Wit S. Uncertainty in predictions of thermal comfort in buildings.Dissertation, TU Delft, June 2001.

    [16] SBIC website, Sustainable Buildings Industry Council, Energy-10.URL: http://www.sbicouncil.org/enTen/index.html.

    [17] TRNSYS Overview website, Solar Energy Laboratory, University ofWisconsinMadinson. URL: http://sel.me.wisc.edu/trnsys/.

    [18] ESP-r website, Energy Systems Research Unit, University ofStrathclyde. URL: http://www.esru.strath.ac.uk/ESP-r.htm.

    [19] EnergyPlus website, US Department of Energy. URL:http://www.eren.doe.gov/buildings/energy tools/energyplus/

    [20] Eastman CH. Building product models: computer environmentssupporting design and construction. Boca Raton: CRC Press,1999.

    Design analysis integration: supporting the selection of energysaving building componentsIntroductionApproachA strategy for the selection of energy saving building componentsUnderlying fundamentalsSystems engineeringEngineering designDecision theory

    A framework for the selection of energy saving building componentsAnalysis of objectives and constraintsDevelopment of an `option space'Determination of the performance of all optionsSelection of the most desirable option

    Support for the selection of energy saving building componentsSpecific needs for support of the selection of energy saving building componentsSupport provided by current building energy analysis toolsSupport to be expected from the solutions resulted from ongoing R&D efforts

    Providing support for the interaction of design and analysis: the Design Analysis Interface (DAI) InitiativeBackgroundGoal

    ConclusionReferences