THE FUNCTICU OF 'IESTINiG ARCHITECTURAL DURING DESICI by GUY FE ARD E. JEINZAPFEL Arch.,, University of Arizona (1965) SUBMITTED IN PARTIAL REQUIFEMENTS FULFILLMIENT FOR MASTEP CF TEE DEGREE CF THE CF ARCHITECTURE at the CF TECHNCICGY MASSACHTSETTS INSTITUTE APR 14 1971 1971 January, LzPARIFs / /'12 , /2 fl/ Signature of the Author Departrn i f Architecture, Janua yQ 71 Certified by . Supervisor Thesis Accepted . ............ 1y Chairman, Departmental . . . . . .a . . . Corriittee on Graduate . . . . Students Document Services Room 14-0551 77 Massachusetts Avenue Cambridge, MA 02139 Ph: 617.253.2800 Email: docs@mit.edu http://libraries.mit.edu/docs DISCLAIMER OF QUALITY Due to the condition of the original material, there are unavoidable flaws in this reproduction. We have made every effort possible to provide you with the best copy available. If you are dissatisfied with this product and find it unusable, please contact Document Services as soon as possible. Thank you. The text on some pages is very faint. This is the best quality available. 2 ABSTb A CT The Function Of Testing During Architectural resign Ey Guy Idward Wcinizapfel Submitted to the Eepartient cf Architecture on January 22, 1921, in partial fulfillment of the reguirements for the Degree of Master of Architecture. The otjective of this study is to explore and clarify the role of testing in the process of architectural design. As the basis of the study, a general definition of testing is developed. Several different kinds of testing are discussed in terms of this definition. oundaries are outlined between testing and the related functicns of problem description and evaluation. And ways in which seemingly useful tests can lead to erroneous conclusicns are discussed. A design experiment is conducted as a means of exploring some of the ways in which a designer employs tests during his search for a problem soluticn. The experiment indicates that tests are made of three important aspects of the design: of the alternatives generated, of the criteria by which the alternatives are judged, and of the design ohjectives which the criteria represent, Based on the experience of the design experiment, a simple ccipputer routine is developed which can aid a designer in testing alternative forms generated ty a computer system. The routine is developed as an example of the way in which a common form of testing can he incorporated into a design aid system, The utility of the testing routine is evaluated in relationship to other kinds of testing which designers employ. It is hoped that this study will be useful to the designer and the design educator by externalizing and providing a tentative structure for a portion of the process which they ccnventionally employ, Further, the study may be useful to thte design methodologist who seeks to develop new tcols to aid the designer in his work. Thesis Supervisors: William 1, Porter Associate Professor of Urban Studies and Planning rlimothy E. Jchnson Assistant Professor of Architecture AC KE CW LED GEM ENT S The author would like to express his gratitude tc: Professor William L. Porter, for his inspiration, guidance, and supervision, especially in the. development of the conceptual aspects of this work, Professor Timothy E. Ochnson, for his assistance, suppcrt, and criticism. thrcughout all phases. of the project, John I. Perkins, with whom many of the concepts toth central and suppcrtive tc. this wcrk were explored, for his assistance during the design study, Professor John R. Myer, for his counsel and ccumEnts, Richard I. Krauss, of Ashley, Myer, Smith, Architects, for providing the architectural prcgram and essential information used in the design study, and those professors Architecture and of interest and work in preparation of this and students in the Departments of Urban Studies and Planning, whose design methods has contributed tc the thesis. TABLE CF CONTENTS Chapter Cne: ....... INTFODUCTICN......... ........ Objectives of tIe Study Testing in Design The Remaining ChaFpters THE STRUCTURE CF TESTING.............. Chapter Two: A General Definition Alternate Forms of Testing Tabular Comparisons Compariscn cf Forms Statements of Preference Non-Translatable Tests Defective Testing Inaccurate Standards Irrelevant Standards Inaccurate Representaticns of Form A DESIGN EXPERTMENT ........... Chapt er Three: The Use of IMACE The Design Problem: A Fire Station The Architectural Program Circulation: A major Design Objective The Model for Generation The Design Prccess Arrangements Tests Changes to the Model New Generation Observations Chapter Your: A COMFUTEF TESTING * ... ROUTINE............. 79 The Purpose of The Rcutine The Test Algorithm The Output Possible Improvements Architectural PrOgrar Fcr A Fire Station P11 Frogram for Computer Aided Testing . .. FOCTNOTES....... BIBLIOC APHY,.. ............... .. * to*. .. .. 112 .,.............. . . .* . . .* .. . . *.. 116 TABIE OF FIGURES FIGURE List of Sraces for the Fire Station FIGU FE Matrix of Generating Specifications ......... 70 FIGUEE Arrangement ......... 71 FIGURE a) Arrangement b) Diagram of Necessary FIGURE 5: a) b) From First Generation Frcm Second Generation ......... 72 ......... 73 Circulation Arrangement Frcn Third Generaticn Diagram of Nccessary ......... 42 Circulation FIGURE 6: Diagram of Desired Circulation ....... ......... 74 FI GUIPE 7: ......... 75 Arrangement From Fourth Generation ... FIGURE 8: Final Computer Generated, Arrangement . .......... ...... FIGURE Designer's Sketch of Alternative 76 Arrangement FIGURE 10: Conceptual Diagram of Solution Spaces...... FIGURE 11: Diagram of IMAGE rata Structure ........... FIGURE 12: Flow Chart of Testing Algorithm. ........... FIGURE 13: TyFical Eistlays of Test Information F IGURE 1 4: Diagram of Possible Data Structure ........ .77 .. 78 ... 0 .. .91 92 ...... ... 93 Chapter One INTECDUCTION This study deals with testing as it is the process of architectural design. performed in It is the result of a study of testing which grew out of a reSEarch effort to develop a computer aid for preliminary architectural design. (1. 1) I. It OBJECTIVES OF THE STUDY the author's hcpe that the ideas put forth is here will help clarify that portion of design dealing with tests. These ideas might also facilitate the develoFmecnt of design aids capable of assisting the designer in testing his alternative solutions. time when much effort is aids for the designer, actual activities, those activities, being devoted At a to developing a clear understanding of his and of the tools he needs to perform is especially necessary. The study was nct undertaken to prescribe any particular process for the designer. -Rather it is the author's belief that no single process could le employed for twc prcblems by one designer, much less by two different designers, The particular functions which are CHAPTER CNE: INTRODUCTICN performed during design, are selected, organi2ed, and performed under the influence of the conte::t of the problem being addressed. It is the problem, the state cf the information concerning that problem, the designer's experience, the resources available to him, and innumerable other factors uhich make it infeasible to prescribe a particular procedure for even the single function of testing. What the author intends instead, is to develop a general discription cf what composes a test, a few of the characteristics of testing, to cutline to provide an example cf the way testing can be employed in design and the impact which it can have on the designer's process, and to develop a simple aid which a designer can use fcr testing. II. TESTItG IN EESIGN This study grew from the realization that testing is a very important function of design and a function which could be analyzed separate from the otler activities of design.(1.2) A few examples of how testing can be used in design may make this point clear. It is often the case that in the process of generating solutions to his problems, a designer develops an alernative arrangement with which he is pleased. He may not be interested in generating new very CHAPTER CNE: IETPCDUCTILCN alternatives but only in ascertaining how well the present alternative satisfies his criginal problem specification. In other words, he would like to test his present arrangement against the criteria cf his original specification. As an example, let us assume that an architect has developed a problem specification for a schocl, and is generating alternative arrangements. Each arrangement may be radically different from the last. At some point, an alternative is produced which appeals to the designer in several vays, The spaces may be organized hierarchically, indicating their functional importance. The overall arrangement might also.be very symmetrical (or asymmetrical) or possess some other order which appeals to the architect, simple structural system, It might lend itself to a and have several other attributes which please the designer for reasons he may or may nct be able tc identify. The architect is reluctant to continue his search for basically different design alternatives. He may be willing to make limited modifications to imprcve the scheme, tut he is very reluctant to make large changes in the 1asic order he sees. It is possible, however, that there is no reascn tc change the present alternative. It specified., may already satisfy the criteria he has He will therefore want to test the design CHAPTER CNE: IETCDUC1C1CN against those criteria, If the scheme docs indeed satisfy them all, be mill. probably accept its lasic form and Legin elabcrating it in more detail. If it fails in small ways, he may rcdify certain rortions of the arrangement until it does meet all the criterea, until he has decided that the arrangement will never satisfy the criteria and rejects it, or until he decides that some of the criteria are not significant enough to warrant discarding tte design and ignores them, In order to pursue his design, then, the architect must be able to test varicus alternatives against the criteria fcr which tLey vere generated. SYNOFTIC In AFPPAISNL another situation, the designer may wish to measure an arrangement against additional criteria, criteria he had not previously specified, Obviously not all criteria can be, or even shculd he, described at the cnset of a problem. Many can be discovered cnly through the process of exploring different alternatives. if all criteria were knoun, unnecessarily complicated it Even would probably le to generate alternative fortcs from the complete set; scue limited subset of the criteria might be all that is acceptable alternatives. the "good" designer is necessary to generate In fact, it can be argued that the designer who selects the CHAPTER CNE: INTRODUCTICN criteria which cause the rcst efficient search fcr a solution. (1.3) As an example, the generating specification for a school may not deal vith the secondary roles which the building is to fulfill. The playing fields might serve as neighborhcod recreaticr axeas, and the physical complex might be used for civic events such as PTA meetings. These functions may not be significant enough for use as generating criteria, lut they are significant enough to be tested in the light of an otherwise acceptalle design. Many times, to light only in seccndary test criteria can be brought the context of a generated alternative. Clients, because they are not versed in the opportunities and limitations they face, often forE new goals upcn seeing the consequences of their criteria. This is even true of the iost sensitive designers. TESTING FCRES CF UNKNCWN CRIGIN A third rcle for testing is typified by the situation in which a designer is design he had no hand in modify to meet a new use. faced with an existing creating, but which he must In this situation, he will want to test how well the design already satisfies the criter ia describing tha t use. Buildings remodellings are obvious examples of this CHAPTER CNE: INTRODUCTICN condition. level, Found object sculpture operates at this ard in nany fields objects are created 'or one purpose which are fcund to satisfy other roles. cases, In these the designer tests the existing forms against new critera, criteria for which they were not originally generated. All of these situations indicate that the designer cannot generate alternative solutions exclusive cf testing - that a very large part of the design process lies outside the realm of generation. However, they do indicate that testing can cccur exclusive of generation. The realization that testing vas a very important function of design and that it could have a "life of its own" separate fror generatiCn, led the writcr to examine the nature of testing architectural form, in the process of design. and its role The results of that study are the substance of this report. III. TEE REMAINING CHAPTERS The remaining chapters can be seen as three different approaches to the subject: as theory, documentary and as application. testing in a general way. as Chapter Two discusses Chapter Three documents an actual design study, and Chapter Four descrites an application of the experience gained from that design study to the development cf a computer aid for testing. CHAPIER CNVE: INTRODUCTICIN In order to establish a common basis for further discussion, Chapter Two, "The Structure of Testing", develops a general definiticn for a test as related to the process of design. It discusses what factors are necessary to form a test, how they are derivEd, how they operate and interact and what results they give. Scme of the different kinds cf tests which a designer performs are discussed in terms cf this general definition. It is shown that both the representation of the design being tested and the state of the ctjective which it is to achieve will affect the way in which the test is performed. In crder to identify the boundaries of testing and to distinguish it frcm other, closely associated activities, brief descriptions are given for measurement and evaluation of design. Cther factors, such as testing's facility and complexity are also treated in an effort to clarify the territory included within the boundaries of this function. The final part cf Chapter Iwo addresses the relationship between the otjcctives which thE design must achieve and the tests which can be made of the proposed scluticn. It outlines the interdependancies between the different eleients of the test, the state of the alternative and the resouices of the designer. And it discusses how faults in those elements can render results which are any of CHAPIEP CVE: IITRODUCTICV inappropriate, mislcading cr ine.ffectual. Chapter Three, "A Design Experiment", documents an actual design study which was performed to illustrate and enlarge upon the issues introduced in Chapter Two. The criteria are outlined for selecting the prcject used in the study, a fire station. Since the design alternatives were generated by a computer system, the reasons for using the comuter, as-opposed to designing in a conventional manner are discussed. documentation of the design process itself The records uhat steps were taken, what tests were performed, and how the results of the tests caused modifications in the designer's strategy. Chapter Four describes "A Ccmputer Testing Foutine" which can be used for testing the alternatives generated by a computer systeu. This routine was developed in response tc the experiences cf the design study. It uses many cf the procedures in a presently existing system. Some of the different kinds of testing are compared to the developed routine, and some prcpcsals are made for expanding the routine to include a wider range thc designer's needs. Chapter Two THE STRUCTUE OF TESTING A test is a comparison Ietwcen the actual value of some aspEct of the design and scne value which it is supposed to achieve. The results of the comparison are acceptatle within certain hirits and unacceptaile beyond those limits. AN EXAMPLE In order to introduce the basic elements ccrposi.ng a test, an example of a typical design test is below. dcscribed Following the example, a general definition of a test is derived, using illustrations from the example to clarify its concepts, Assume an architect is designing an office building for a developer. One of his client's primary objectives is that the building should make as much profit as possible. Knowing that the ratio of rentable to non-rentable floor area vill significantly influence the ultimate profit return of the building, the architect and client seek a ratio which will give the greatest probability of a high return. this ratio is They know that rot the cnly factor bearing on profit, but CHAPTER TW1O: that it is ST(RUCTUP significanit enough to le carcfully considered. From his past experience, the architect has found that fev buildings can he expected to achieve a ratio greater than six to one, least four to one. but that most are capable of at The client's experience in real estate developuent shows that buildings with a ratic less than 3.5 tc 1 will return a marginal profit even when fully occupied, and that those with greater than 5.5 to 1 fail to do much Letter because of high vacancy factors caused by the teniants' like sardines". around 4.5 to 1. feelings of being "packed The optimum return appears to occur They decide, then, to seek a design which will have a 4.5 ratic. They will be satisfied with nothing less than four, and will avoid alternatives higher than five and a half. In seeking a basic arrangement for the building, the architect generates several alternative scherres. Each shows the basic configuration of spaces In the building. Corridors, offices, elevators and nechanical equipmert rooms are all rcughed in. He determines the area ratics for each scheme by the process of subtracting the total rentable area from the gross area. The ratios of each alternative scheme are ccmpared to the standard he hopes to achieve. Those scLemes with ratios greater than 5.5 cr below 4 are set aside. Some, CiAP'IFR TWO: SURUCTUPE in fact, are rejected. Cthers are modified in hopes of meeting the standard. In accepting or rejecting the alternative schemes, the architect is tests. basing his judgcment on the results of These tests are typical and embody the general form and the basic ccnditicns of testing. discussion in To clarify the renainder of this study, definiticn and notation is 1. a specific given below, A GENEFAL DEFINITION A test, T, may he defined as a comparitive function, t, Lctween sorte neasurement, and a measurement, acceptatle M2, of an objective, within scre tclerance, T = f ( N 1(F) THE FORM, F, is M2 (0) which is t) the description or the the fcror is the office building. 0, F, t. representation of a pcssible real form. example above, M1, of a form, It is In the test the alternative scheme for actually a representation of a possible form that might ultimately be built and not the reality of the building itself. For some tests the form may also be thought of as the larger environment within which the building would be only a part. necessary in This larger conceptual frame is order to test those aspects of the design CHAPTIER TWC: STiFUCTUFE which lic partially cutside the description building being designed. would test As an example, the solar exposure of surrcunding buildings, a designer of a building, to have information concerning of the who wculd have the location and shape the range of positions which the sun would follow over tlhe year, the climate of the region, as well as the crientation and shape of the building itself. THE CEJECTIVEj, C, is the goal for which the environnent is being tested. normative in nature. Unlike the form, it is It states how the form should exist rather than how it does exist. In the example, the objective states that the build.ing "should make as much prcfit as possible." M1 and M2 are the functions which measure the form and the objective, respectively. They are noted differently to indicate that their measurement techniqucs and the data they rely upon are not necessarily the same. of the form is In the example, the actual ratio deterLTined frce a function which scales the varicus dimensions of the form and by mathmatical operations computes the actual ratio of the proposal. The objective ratio derives from the past experiences of the client and the architect. The reasurement of the form derives an attribute, "a", some value which nay be attributed to the form. CHAPTER TWO: SIRUCTURE M1 (F) = "a" The measurement of the objective standard, "s", which the forn = M2(C) In 4.5. the test The exanple, derives a should achieve. "s" the value of the standard value of the attribute varied was for each alternative form. THE 1GLEBANCE, between t, is the allowable deviation the attribute and the standard. It is the amount of misfit which the designer feels he bet;een the design and its objective. In can accept the cxample, the tolerance permitted a range one unit higher and one half unit lower than the optimum ratio of 4.-. The tolerance is perhaps the most flexible aspect Nany tests uill maintain of the test, the same objective and standard throughout a design, tolerance is varied while the from very loose to very specific in keeping with the detail of the alternatives developed. For example, the main entry to a building may at first be acceptable anywhere cr the north facade. arrangemEnt of the building's interior, Eut as the and its immediate environment, beccues mcre specific, there will likely be only one acceptable location for the door. The test operations. function, It first f, is composed of two compares the attribute and the value cf the standard, of the form thereby deriving a third CHAPTER TC: STRUCTUEE measure, the difference LEtLcen the two. then compares that diffeiarce The function with the tolerai.ce, t, and passes or fails the test cn the basis of that Thus, the form of a test. is: 1) the comparison. measurement of some aspect of the proposed fcrim, r1 (F), which results in a value of that aspect or attribute, "a"; 2) the measurement cf that same aspect of the objective, M2 (0) , which results in a value for a "s"; 3) the ccmparison of those two results, standard, "a'and "s", resulting in 4) a measure of their difference; the ccmparison of that difference with the tclerable difference, and 5) t; the passage or failure cf the test derived from that final ccmp.arison. II. ALTERNATE TORMS OF TESTING Not all testing is the rathematical comparison of quantifiable values, above. such as illustrated in the example many different kinds of tests are possible, dependant upon the state of the form being tested and the type of objective it is to achieve. That is to say that tests differ in kind due to the varied operations and procEdures they require. Diverse forms of measurements are required by the many modes in which a form may be represented: as a mental image, as a verbal descrirtion, as a drawing, as a model or as a final CHAPTER TUC: SIRUCTUEE constructed reality. standard, performed, The type of cbjective, and its vill also influence the kind of test Many objectives, such as lighting till necessitate reference tc some pre-established table for quantifiable standards; footcandles". "The light level should be 75 Others 'will require an indication of qualitative preference, to externalize, Yet others will be impossible or translate into any explicit form. In short, different kinds of testing are produCEd by the test's two rajcr arguments and the procedures used to measure and comrpare them. Scme cf the different kinds of tests pcssible are discussed below. TABULAR COMIPARTSONS A common type of test'is the comparison cf sone aspect of a design to some tabulated or quantifiable standard. In tests cf this nature, the standard is derived frcm a table, where M2 is the search of that Ixamples table, or the recall of its previous search. abound as entire volumes have been devoted to iecording such factors as hearth/flue area ratios. (2.1) nearly every profession and trade, And as well as Uost federal, state, and city bcard, regardless of how remotely connected with building ccnstruction, produce copious guidelines for the designer. An extension of this kind of testing is made by substituting an equation frcm which the standard can be CHAPTFR TWO: STRUCTURE calculated. Rules cf thumb, such as "The riser plus twice the tread should equal 29 inches", are more convenient, yet produce the same effect as tables. The derivation of structural and mechanical equipmcnt standards are extensicns cf this approach. Tests of this type are easily described, usually tased on mathematical calculations and comparisons. And the tests can Their results are also quite precise. be performed by even the most superficially informed. CCEPARISCN CF FORMS Another ccmmon kind of testing employs the compariscn of two forms. In this case, the standard is intrinsic within ancther fcrm or its which is desirable representation known to be acceptable or to have certain characteristics. The test is based upcn a comparison betveen the "otjective" form and the form under question. For exaiple, a designer might develop a scheme in which a very cbvicus hierachical circulation system is evident. He can test his other alternatives against this scheme to determine if such clear circulaticn had been generated unnoticed before cr to see if the other forms cculd be easily modified to incorporate it. By an extension of the notation above, such a test can be represented as: T=f (M1(F 1) M1 (F2) t) CHAPTER TUWO: where F1 is SRUICTURE 22 the alternative being tested, objective forw (in this case, clear circulation), F2 is the the scheme having the and M1 is a common process of measuring the twc forms The objective form, (here a visual inspection). F2, need not be otherwise related to the design protlem for which the test is conducted. It Kight as easily be an existing building, or an alstracted pattern, wheel. such as a grid, a tree, or a It need only erbcdy or represent the objective being tested. representations Diagrams are commonly used as simple of fcrm cbjectives, A frequent variation of this kind of test uses the objective form as a "threshold standard", of a least acceptable level, some statement which may be cou-bined vith an indicated directicn for improvement: "The facade should he at least as impcsing as Bonwits - hopefully more so "; or "This building must be at least cne story higher than the Prudential." Obviously thresIcld standards are applicable to other kinds of tests as well. STATEMENTS OF UEFERYNCE Most standards Pay be seen as statements of someone's preference. Certainly the tabulated standards CHAPTER TWO: SqRUCTURE for heating and lighting mere determined by the careful synthesis of many pecples preferences. Hovevcr, Eost standards are not primarily identified with this characteristic. But certain test criteria gain their identity solely from the objectives cf some actor central to the design problem, the client, the rser, the designer, etc. The "Observations" in the Fire House Architectural Program in the Appendix contain many examples of preference statements; "There will be a flagpole." Many prefernces can not be foreseen pricr tc the generation of alternatives. As a result, many tests in this categcry arc perfcrmed by using the design alternative to extract the preference of those involved. In such cases, the distinction betueen the elements of the test become blurred, for in a sense, the alternative is measuring the clients desires, which in turn pass or fail the scheme. NON-TRAES ATAELE TESTS At the opposite extreme from tabular comparisons are tests in which the standard, and even the objective it represents, cannot be made explicit, and in which certain aspects of the alternative cannot be universally understocd. A designer nay continue to search for alternatives while nct being able to explain his reasons for rejecting those at hand. CHAPTER T1O: S'RUCTURE It is perfectly understandatle that certain design goals will be inexplicit, image cf siice the designer's his problem is very rich and derives from many diverse sources. His prcblem image goes well beyond the objectives stated or inferred by his client to encompass his grasp of technological, economic, and political systems - their resources, limitations, and potentials, Further, the designer's image of his problem is the culture and society cf which he is based in a product, impacted by his previous design experience. and is These and many other factors ccmbire to form a very rich perception of his problem. Moreover, both the nedium in which the design is addressed and the discrepancies letween that redium and the consequent reality cf the built form pose issues which cannot be externalized communication. in all forms of The homily, "One picture is worth a thousand words", is not entirely accurate; words will never recreate the picture. In a similar way, many of the attributes and relaticnships ccnveyed by representations of the design cannot be communicated adequately in any other medium. Ccnversely, because of his familiarity with both his. mede of representaticn the buildings which result, information and a designer may sense frcm his representation of a design which is not grasped by others. CHAPTER TWO: SSIUCTURE As a consequence of the designer's diverse perception of his prcblen aInd the iedium in many explored, tests will be conducted of externalizing possibility their results. vhici it is without the prccess or either their As opposed to tabular comparisons, tests of this kind can be performed only by the designer, himself. AN ELEMENT CI TESTING MEASUREMFNT: For many, a certain anbiguity exists regarding the meaning of the word "test". Tresting is often used to refer to functions which are not actually conmrarative in nature. For example, determining the height of a sp-ace or the numter of intersecticns on a corridor are considered by some tc he tests the design. testing the state of For purposes of this paper, however, determining the state of some aspect of a design will be refered to as a measurement, or a description. when informaticn concerning the design is comparison Only usEd for a with an cbjective can the function le identified as a test. TESTING: AN ELIMET OF EVALUATION A similar confusion exists regarding the distinction test is between testing and evaluation. whereas a the .compariscn between the actual state of a certain design alternative and the state necessary to CHAPTER TWO: STRUCTURE satisfy a single objective, evaluation is taken to rean either 1) the assessirent cf a design alternative's worth tased upcu the ccmparison of that alternative to a co2mprehensive set of objectives for the project, or 2) the comparison and ranking cf several design alternatives according tc their relative satisfaction of a set cf criteria. In both forms, evaluation must determine the relative significance of conflicting goals based upon the preferences (the value system) of some person or persons, ic. client, user, designer, etc. Evaluation assigns the worth of a particular design alternative; it ranks the relative "gocdncess" of several possible designs. In doing so, evaluation may employ testing in much the same way that testing uses measurement. But evaluation poses conceptual and procedural guestions which are well beyond the scope of testing and this study. HE FACIlITY OF TESTING None of the conditicns which determine certain functicn is or evaluation) that a a test (as opposed to a measurement have anything to do with the degree of difficulty involved in involve as little performing them. A test can effort as a simlple visual inspection or can require a complex and arduous mathematical calculaticn. Perhaps most design tests are the result CIIAPTER TWO: of brief is STRUCTURE visual observation, since the state usually described by a drawing cr model. cf a design Put the detailed ocinputations neccssary to determine an expected foot candle level and compare it to the intended standard do not make that function anything micre or less than a test. IIT. DE!ECTTVE UESTS Several factors can cause tests to be inneffective, thus leading the designer to errcneous conclusions. major elements of a test, and the derivation Two the description of the form of the standard, influence thc effectiveness of the test. This section will discuss: the ways in results, which inaccurate standards can cause useless the relationship Letween irrelevant standards and meaningless results, and the limitations cf tests which measure a rcprcseta ticn of a form rather than the actual form. INIACCURAIF If STANDARDS the value of a certain standard against which it passed, If the result, the standard is attribute natches the Yeing compared, itself is tle test is inaccurate, however, and hence the test may be meaningless. CIIAPTER TWO: STR[UCTURE Standards should be determined which accurately represent the higher level cbjectives for wihich the test is being conducted, to Any standard mhich fails accurately represent its cbjective vill cause misleading results. For example, a design may be sought which facilitates easy installation of plumbing. Chases of less than 6 inches may, in actuality, be less than the required clearances cf scme fittings and thwart that objective. A standard of 5 inches vould be inaccurate and tests hased on that standard would be misleading. There are many reasons for which standards might be inaccurately derived, Among them are: are not easily quantified; 1) standards which 2)standards for which no universal value has been set, either because the objective is unique cr because insufficient study has teen done to codify them; 3) standards uhich have become obsolete. Many standards are especially difficult to quantify. Qualitative objectives often fall in t1is group. Privacy, openness, s pacicusness, cheerfulness are a few of the goals with which an architect might be faced. in They convey attributes which are very important any design. Unfortunately, subject to interpretation. designs for his home, their value or ireaning is A client, seeing the first right reali7e that his conception of privacy and that of his architect are quite IWO: STRUCICTrPE CHAPTIER different. The designer must szupect the standards Ye right estatlish on the basis of qualitative objectives. Tn such cases, however, he may be aided by a special By propcsing a range of design application of testing. alternatives to his clients, the architect can elicit a response as to how well each achieves privacy or In this way, he can begin to get a spaciousness. picture of his preferences, and be able to mcre accurately place the value cf the standards. Sore standards, though quantifiable, are difficult to measure accurately in the absence of the pecple for whom they are created. Cften, the -designer cannot directly contact the people for whom he is these cases, preferences. avoid. In designing, he will lack the information regarding Inaccurate standards will be difficult to An excellent cxamplo of such a case occurred at the Seattle World's Fair. Several of the science displays at the fair were very complex and required abundant time and effort to be understocd. It visiter examined was chserved, however, that the average the displays fcr no more than a minute, and that very few visitors remained for more than five minutes. When asked why the displays had been so elaborately contrived, the designers responded that the intent was not to attract only the casual fair gcer, but CHAPTER TI-C: STRUCTUE also the serious student cf the subject being displayed. The designer explained that the display had been organised to offer information to such a student for as long as an hour. The designer had ccnjectured the existence of the serious student, and tested the display's content on the basis of his existence. attented the fair, In fact, no such student and the displays were too ccwplex for the average visitcr vhc did come. In effect, the standards which tle designer had met were inaccurate. (2.2) Several techniques are available which car assist the designcr when standards must he set in tc alsence of those for whom they are created. They may he predicted by the designer frcm his own experience or from his empathy with the user, shortcorrings of conjecture But the possible are seen in the exanple cf the fair displays. Experinents of pilot situaticns, and case studies of existirg, analogous situations can be very useful. detail in These techniques have been covered in numerous sources, and are beyond the scope of this study. Inaccurate standards also occur because shifts in preference make them obsolete. which have been very accurately In many cases stardards determined will over long periods of time !(in time scales larger tian the CUIAPTER T"C: STBUCTUFE schcdule fcr any design prcject) change and invalidate their usefulness. Fcr examrle, were accurate twenty years housing standards which ago are no longer adequate today. Even within the time scale of a design, client preferences may alter enough tc invalidate certain standards, This may cccur out of the resolution of conflicts between standards or because of the appearance of new alternatives. IRRELEVPNT STANDARDS Standards may be accurately measured, hut they Pay be irrelevant to the objective sought; they ray te incomplete, failing to acccunt for additional, dependant standards; they tray also be inadequate, failing to account for all the factors they influence. Standards, context, though accurately measured for one may be irrelevant to the objectives they serve. As an example, low income housing guidelines have been criticized as irrelevant for the people they serve. professed objective cf the guidelines is The to insure adequate housing for the needs of low income families. They are developed on the tasis of experiences with middle class farilies, and set minimum standards for living rcoms dining rocrs, Unfortunately, kitchens, kitchens and so-forth. dining and living rooms fail to CHAPTER TVO: SI PUCTUPE reflect the actual life styles which lower inccmie The standards are irrlevant tc the families pursue. needs of the users, living room is not sitply inaccurate. Whether the 500 or 6CC square feet actually makes no difference, The spatial centex t of the standards does not "fit" the life style of the users. Lack of sufficient qualifications can also irake an otherwise meaningful standard irrelevant to its objective. Many standards do not stand alone. They are actually cnly cne factor cf networks cf interrelated criteria. useful. Alone they may seem very realistic and they may even give reasonable results when considered alone, if their dependant criteria happen to be satisfied as well. the architect is Hosever, ccnditions change and not aware cf one standard's dependancy upon others, the results cf any number of tests can become invalid. Numerous examples of interdependant criteria exist.. Take for example, the "clinate" of a school, where a good environment for teaching is sought. The standard for the heating and air conditioning might require that the tempcraturc be capable cf maintaining 72 degrees in the room at all tirmes. 5his seems reasonable since many experiments have shown that a temperature of 72 degrees is most universally sat.isfactory The building is designed and the hea ting systein meets the tetrpcrature ClIAPTEIR TWC: standard. STRUCTUFE the students and teachers Unfortunately, complain about the heat. fluctuating humidity. satisfactory, met. This is 1Ihe answer turns out to be the The environment failed to be even though the temperature criterion was a very common situaticn one standard being highly dependant upon anctler. Unless the complete network cf interrlated criteria are all satisfied, no number of successful tests on one criteria will produce meaningful results. In these situations, standards is the network of intErdependant best stated as a "ccmplex" criterion. Perhaps even a range over uhich the set is vill satisfactory be necessary to insure effective results. The temperature and humidity relationships might actually indicate a range of catisfacticn from 80degrees and 15 per cent to 65 degrees and 80 per cent. Just as several standards might be interdependant for the satisfaction of a certain criteria, so also might those same standards relate to other criteria which would further ccnstrain them. temperature/hufidity For example, standard necessary, above, the may be satisfactory fcr ccmfort and therefore insure a good climate for learning, But good health might also be strongly related to a certain range of humidity, while the temperature range for laboratory conditicrs right be additionally restrictive. Therefore, the designcr would CHAPTER TEIC: SIEUCTUFF not only have to develop a range of possible tempcrature/huiidity standards for good learning but would have to check their relevance against the criteria of good health and acceptable laboratory conditicns, as uell. INACCURATE REPRESENTATICNS CF FORM The two sections above discussed the ways in which invalid standards can cause invalid tests. Sinilar problems occur from the limitations of measuring representations of the form rather than the actual forms themselves. Several factors make it necessary to design buildings by means of drawiings, representations of reality. the resources required factors make it models and other The size of the prcjects, to build them, and several other infeasible to design most buildings in ways similar tc painting and sculpture. The restrictions imposed are often severe. The conditions which will actually occur in the real building can only be estimated and assuned in design. its The uncertain nature of these assumptions may cause severe inaccuracies in design, wind forces for which a building is For exanple, the designed are only assumptions. The building can be designed to withstand these loads, but in actual fact the winds ray te CHAPTER TWC: STRUCTUBE substantially higher and cause the building to fail. It is because of just such uncertainties between reality and its representation that safety factors are set so high. Limitations in accuracy are also due to discrepancies between the designer's specifications for the building and the constructed reality in its final form. Highly specific, very cautiously prescrited representations will always differ from their final For example, realizations. 10'-4 1/2" floor to floor dimensicns may be exactly specified in documents. the contract When the building is constructed, however, the dimension vill certainly differ, from that reasure to some extent, But often, Cften the difference is as in inconseguential. structural design, the variance is quite critical. Measurements of representations may be invalid for reasons cther than the descrepancies between designed and built -values. They may be incomplete as well. representation, short of the actual artifact it models, contains only a limited descripticn of the fcru. is true by definition. informaticn result, Every This If it contained all the of the forn it '%culd be the form. As a measurements cf a representation often exclude significant information. And tests based on incomplete measurements may lead to invalid results. Unknown CHAPTEP TWO: SIRUCTUR numbers of architects have designEd doors vhich, the plan, appeared tc orer freely, frcr but which were actually constricted by overhangs or elevated projections. ests for adequate egress were satisfied by the information taken from the plan, but they were most invalid in reality. One architect in the Scuthwest was disappcinted during the construction cf a department store he had designed. He found that a gusset plate which supported a mezzbnine girder passed six inches into the clearway of an elevator shaft. All his checks of the drawings had teen insufficient tc show the situation which existed in reality. Many qualities embodied in the form of an actual building are more than the sun of their quantifiable measures. And representation of the building are inadequate to convey these qualities. The reverberation time of a concert hall can be clcsely predicted fron the drawincs. The background ncise level can be determined. The absence of echoes and flutters can be virtually assured. In sum, all the quantifiable factors for good hearing can be validly tested. And yet the ball nay still be an acoustical disappointment. the the result in more than one case. failure is in This has been This type of not due tc uncertain assumptions, the cCnstruction. or faults Rather, this type of failure CHAPTER TVO: S RCTU E 37 37 results from qualities which cannot be conveyed in the abstract. SUMMARY The discussion above points out many of th e ways in which the results of tests Eight be invalid. is not ccmplete, but it TIe list dces indicate a few of the broader catagories of failure: inaccurate, irrelevant and inadequate standards as well as inaccurate and incomplete representaticns of reality. The designer, in order to conduct a test, nust actually test the validity of the parts of the test and be ccgnizant of their limitations. derived, to detect He must understand how his stardard was what data it was based upon. He should be able those norms which are inaccurately reasured or which have heccrfe cbsoclete, either because of changes in his design objectives or because of changes at levels different from his design project. He should know which standards stand alone and which are valid only in the . context of others. The designer should have a sense for the appropriateness cf a standard know how well it to his objectives and reflects the many factors it may affect. Further, the designer must he alert to thc possibility of significant discrepancies between the representation of his design and its subsequent reality. To be aware of these factcrs and their significance, the CHAPTER TTWO: STPUCTURE 38 designer must constantly mcnitor the arguments of his tests during the design piccess and review tIe validity of their ccnseguences in building. the final reality of the Chapter Three A EESIGN EXPERIMENT In order to illustrate and expand upon the issues discussed in problem, tie preceding chapter, a fire staticn, a small design kas studied. A case study approach was used for this experiment in order to explore how tests are actually performed, and how several interdependant tests are often necessary to provide adequate analysis cf a given objective. The case study was also used to explcre the ways in which tests affect the design process, and to clarity what kinds of tests, and what form of results are useful to designers. Alternate design soluticns were generatEd for the problem by the ccmputer design aid system, IMAGE.(See footnote 1.1) These design alternatives here tested by both conventional and computer aided processes. Some cbservaticns regarding the kinds of tests made during the design study are giver at the end of the chapter. THE USE CF IMAGE The experinent was conducted system for several rcascns. with aid of the IMAGE First of all, because of C1APTER TEREE: EXPERILEN'I the nature of the computeL system, the different steps Since taken during the design uculd be very explicit. the computer system cperates on a very specific, though limited, set of infcrcaticn, and since it to generate form alternatives, operates only both the state of the information and the design functicns being performed would be clearly kncun at all times. Since IMAGE had no testing capabilities at the onset of the study, conventicnal manner, alternatives. the tests would be performed in a separate frcm the generation of the This vould further isolate the testing function and prevent its confusion with other operaticns. It vas hoped that some p1orticn of the tests could be computerized during the later stages of the experiment - that the experilent would help identify those tests which were mcst suitable for computer applications, either because of their recurrance, the. designer's need for accurate results, or because of their difficulty. As it happened, the study did point out a need for a limited testing routine whic1 led to the development of tLe Ircccdure described in Chapter four. 1. THE DESIGN PPCELEMN: A FIRE STATION A fire station for Boston was selected for the CHAPTER THREE: EXPERIEEIT was a relatively well defined study because it architectural understood. problem, whose majcr goals were clearly Those goals, of course, were to provide housing for a squad of firemen and to provide maximum egress from all parts of the housing to the fire fighting apparatus. The cfficial objectives and requireaents of the prcbleu vere set forth in an architectural space program which was standard for all fire stations in the Boston area. A list of spaces specified in the program is given in Figure 1 cn the follcwing page. in The architectrural program is included the appendix. (3.1) The fire station vas also selected because it was a relatively small problem, and a problem for which there appeared to be only a few acceptable design alternatives. spaces, That is, because of the limited number of and the emphasis cf a single function, only a limited number of basic arrangements could be expected This made the fire station a to satisfy the problem. particularly good problem with which to study testing, since the standards of "good fit" between design and objectives could be clearly described. THE ARCHITECTURAL PROGRAM Though the architectural prcgram was typical of those used for the design cf all Boston, and though it fire stations in was considered to be fairly FIGURE 1: LIST OF SPACES ICE, TIRE STATION Area Function Apparatus Room,.... PatrLCl RCC * .. .. . .. .. . . b*5@e)5*J S 120 .... 25 ... Washroom.......... Hose trying Tower Hose Store Room Dousing Shower Clothes Drying Foom........... ... Oil Stove Gencratcr Rcom Boiler and Mechanical Foom Provisions for Air CcmFressor. '.., Engine Dormatory....., 2,800 ,...... %..... 3CC 5.... 545 * 700 g55* ... ladder Firemens' Washroom... ..... Firemens' Shower Locker .. 5** 700 I... *9 ~55 150 '.. 0e594 1510 .5... 650 .5... 315 08550 315 eggs. 250 Dormatory...,,,.,..,.., .. Room....*... ec............... Engine Officer's Rccr... 1 *.. Ladder Officer's Rcc........ Officers' ... Viash and Shower Room S ... Kitchen-Dining Recreation Focm........... room............ C C C 4~ S. SC Linen Closet Stationery Store Janitorial Cleaning Closet Household-Utili ty Surplies closet and Cleaning Storc,.......... Janitorial Two Fuel ranks Gascline Fump Frcnt Aprcn Parking for 25 Cars 320 480 205 CHAPTER THREE: EXPERIEEN7 comprehensive, form goals, it 43 specified only a minimur numler of mcst of uhich dealt uith room sizes and adjacencies. While the activities planned for the different rooms were labelled, they were not elaborated However, room furnishings, equipment and upon. mechanical services were outlined in some detail. Optimal or satisficing dinensicns were specified only for equipment areas, such as the Apparatus (fire engine) Qualities were described in Room. equipment reguirements, terms of mechanical maintenance characteristics, water drainage, and heating/ventilating requirrents. Cnly eight form-specific requirements were actually given, though many cther formal relationships could be implied or were taken fcr granted. The needs of the fire fighting apparatus iere in many ways described more elaborately than thcse of the firemen. Two scale drawings showEd the turning radius of the vehicles and the dimensions to be allowed for their parking. Several reccomendations were given for the sizing, prctecticn and operaticn of the Apparatus Room doors. The typical preoccupation with machinery and its progra!., operation was evident throughout the fire house while the physical, social, and psychological needs of the men using the fire house were seldom directly addressed. (3.2) The lack of comprehensive documentation of CHAPTE THpEE: EXPqRPEENT functional reguirements is not unusual. When compared to most architectural space programs, documentation of the fire staticn is one of the iore complete. And, it was for the abundance of this documentation that the problem was selected. (3.3) Voids in the fire station prograuc point cut the necessity for the designer to acquire additicnal information on his own. This information can come from many sources: the architect's past experience, conferences case, documentation of similar projects, with the cliert, and sc forth. In this most of the detailed irfornation relating to the individual activitics and their settings was obtained through exeiisive discussioins with. the fire department personnel and cbservaticn existing staticns, of opeLations at several This iiformation was gathered by achitects who had designed a staticn, and vas made available for this study along with the architectural program.(See footnote 3.1) CIRCUIAIICN: A MAJOR LESICN OBJECTIVE Because of its impcortance to the success of the fire station, circulaticn was identified as a primary objective of the problem. Unfortunately, the architectural program's description of circulaticn was typical cf the program as a whole; it was sparse at test. cases. Circulaticn was mentioned explicitly in orly ten Tlese references 4ere usually general in CHAPTER THREE: character: EXPERIMEiNT "A twc story building lends itself better to rapid alarm generated circclation than a single story building." The few detailed specifications dealt more with maintenance than uith cperation: "Floors should not be waxed as firemen in a hurry, moving quickly, and dressing as they travel are in danger of slipping." Despite the general reccwmendation that "emergency routes to the vehicles be direct and free frow obstructions", no explicit description was given of which areas those emergency routes should and should not connect. And while at least some attention was given to emergency circulation, use, ncne vas given to non--emergency such as between the dcrmitories and recreation room. Though the criteria of good circulation lad not teen adEquately specified in the architectural program, the additional information Fassed on by the architects cf the prcject and available in architectural journals was more than sufficient to make the requirements clear. Most importantly, no two major routes of emergency circulation could crcss cre ancther. This was true for the apparatus as well as the firemen. A path in front of a fire engine wculd be ridiculous, Secondly, paths from any of the major spaces to the apparatus should be as direct as possible; it any room tc the Apparatus should be possible tc get from ccm vithout going through CHAPTEF TLEE: EXPEEIMENT 416 more than cne cther space. Pequirements for non-emergency circulation were much less important. It was desirable that the officers' rooms be accessable without passing through the firemens' guarters. vas also necessary tc move It easily from the pumping engine to the Hose Tower for Public handling, drying and stcring the hoses. circulation was to enter the station in clear view of the Patrol Room and te kept from penetrating the staticn was seen as the major public too deeply. The Restrocc destination (after the Patrol Rocm) , and was therefore The Fecreation to be kept to the frcnt of the station, and Kitchen-Dining areas were to be equally accessable for both the ladder and engine crews, This factcr could be judged by the distance and complexity of the paths from each group's guarters. Several of its general characteristics made circulaticn an especially suitable objective for emphasis in the design experiment. convenient circulation is architectural probleas. enough, an objective common to all also ccmplex Circulaticn is embracing a variety cf components and characteristics, factor, The need for No single to require many tests. regardless of how satisfactory it is tested, guarantee the overall acceptability of a circulaticn system. Eather, a designer must test his circulaticn can CHAPTER THEEE: EXPEBIMENT against a wide range of interdependant factors: the length of paths, the sequence of spaces they join or exclude, as well as the interest, clarity, directness of the circulation network. factors could, in turn, many other criteria. and Each of these depend on the satisfaction cf Clarity and directness, for example, night relate to the number of interconnections between Faths and the number of turns within the various It sub-elements. was felt, therefore, that these characteristics of circulation would foster different kinds of tests, thereby eypanding the concepts of Chapter Twc. II. THE MODEL FOR GENERATION Alternative forms for the fire station were generatEd by IFAGE frco "model". The initial a limited problem descripticn or model cf the problem was limited to only a simplified set of spaces, program, reduced from the and a very minimal set of relationships between those spaces. It was the designer's intentin that the model be limited for the initial phase of generaticn, and elaborated as more was learned from the arrangements generated. In this way, IMAGE could be used by the designer in an heuristic search for an adequate problem description and solution. It was also felt that by CHAPIER TEREE: EXPE1i11'EINT under-constraining the prcllem, the tests peifcrued could play a greater rcle in the development cf the model. ThE seventeen in spaces used for the model are listed Figure 2 on the following given the same initial page. All the spaces were location and the same size (cne foot square) to eliinate any form preconceptions vhich the designer might have had. specified The relationships between the spaces are also shown in Figure 2. These consisted priuarily of relative size and proportion along with a fev adjacency or proiffity This reflected the architectural tequirements. program's primary concerns. Each space was scaled to a (usually the Apparatus Room) in terms of control space maximum area and a range of acceptable proportions, The intended effect was to maintain specific areas while allowing the location and proportions to vary as widely as possible. Visual access was required from the Tatrol. Room to the Apron, the Apparatus Room, Cf course, and the Restroc, non-overlap was specified between all the spaces, since each was to represent an individual rcom. The model was specified with all spaces on the same level. This was done because of limitations within IMAGE. The system could have dealt with a multi-floor specification, but much time would have been necessary to improve the system. Since this was not relevant to CIIAP7E1 THEEE: the study, EXPFE3BENTI prcgrau did not and since the architectural was strongly demand a two flcor schene (though it suggested), -a single floor model was specified. The primary Effect of this was to further limit the number of possible satisfactory alternatives. The model was additionally removed from reality by the absence of an actual site and its inherent The crly site characteristic vhich was restrictions. built into the model was achieved by fixing the Apron. Since other spaces were related in some way to the Apparatus Boom, this had the effc-t of creating an exclusive "street" cn the side of the Aprcn opposite the Apparatus Room,. But no site boundaries, street were specified, other than the or irillied. III. THf EESIGN PEOCESS The fire staticn study involved five major series of interactions. Each was originated by the generation- of a series of arrangements. The arrangements were tested, and on the basis of the- tests' results, changes were made to tle' model and the next sequence was begun. The designer referred tc during this experiment is the autLor. It would have been possible to use someone else as the designer, thereby avoiding the possibility of a biased analysis of the process. However, it was prohablc that this alternative methodology would have CHAPTER TEPEE: EXPEFIMEET merely substituted cne perscn's prejudices for another. It was alsc possible that rany of the designer'z insights into his process would have been blurred by their double translation, first by the designer himself Lue primarily and then by the analyst. fears, it to these two was decided to ccndense the two roles of designer and analyst into a single actor and to accept which that methodclogy the liuitations implied. 'TBE INITIAL GEEEEATICN Using 2, the generating THAGE enlarged all the spaces Figure from their criginal one square foot area to the size specified. also disaggregated from in ncdel as illustrated The slaces were their single locale ard repositioned in an effort to satisfy the specifiEd relationships. this first Figure 3 show a typical arrangement from series of geneiaticns. ERRORS IN THE ARRANGEM1ENI There were obvious faults which were conrcn to all the alternatives, and which could be deduced easily frcm tests made by simple cbseivation. small by one half. It The Apron uas too was tc have the same dirension on both sides as the Apparatus Room had on its The Cfficers Quarters (E-CF, away from the main cluster way. L-OF, and W-CF) of spaces in short side. were moving an unpreditable But the most chvicus flaw was the fact that the CHAPIEB 1HJFF: EXPEFJIENT 51 spaces were ba-dly cverlapjcd.(3. 4) All three of these conditions were so obvious as to Cnly a minimal understanding virtually test themselves. of the space program and a quick examination cf the display was necessary to see the problems. The errors led to a recheck of the model which was used for the generation. It was found that the non-overlap specifications had been made as interded. Evidently they had to be more highly weighted in to accoffplish their cbjectives. The Officers Quarters had alsc been specified as intended, them tc flcat. crder but this allowed A freedcm which was not desired. The cily specificaticn which had been made incorrectly was the size cf the Apron. This was a simple oversight and easily rectified. It wculd have been irrelevant to test any of the other elerents of these arrangements, since the three errors mentioned were sc severe as to confuse the rest of the design. Proximity was specified between the Officers Washroom (9-CF) and the Firemens Washroom (WASE), and the system was run again. THE SECCNED SET CF AUDANGEMEVIS The best arrangement of the second series of generaticns is shown in ligure 4a. Two major kinds of CHAPIER TIEREE: EXPEBIETI tests were made of this arrangement. First, all the specifications of the imodel were checked for satisfaction. And second, the circulation nCcessitated by the arrangement was analysed. TESTS OF THE ECLEI The model was tested in order to know when it approaching sclutior. when nearing solution, it was was generally not necessary to continue generating because the trend of the arrangements could be seen. Eut if the model was not near solution, significantly different arrangements could still be expected. Frcm tests of these arrangements, it appeared that the generating model had been satisfied exception of three conflicts: with the the Officers Washroom was undersized; and the P'echanical Pocm was not adjacent to either tIe Hlcse Tower cr the Locker Room. Tests of the mcdel were performed by comparing the informaticn crntained in the specification matrix with the graphic display cf the arrangement. Since all the specifications of the model were simple geometric conditicrs, visual inspecticn was usually sufficient to determine their state of satisfaction. Officers Washrcor, fcr exarple, The size of the vas clearly too small. The matrix showed that that space was supposed to be 250 square feet wiile tec Washrccm (EASH) was to be 150 s.f. Since the display showed t-OF to be smaller than WASH, CHAP T ER IEFEE: EXPERl]LENI 53 cne or perhaps both were incorrect. Compariscn of the WASH to cther sraces led tc the ccnclusion that V-OF was the odd space, probably having been shrunk by IMAGE to alleviate cverlaps. This.kind of multiple ccmparison is of testing. a ccmmon form Many designers draw design sketches entirely by proportional relationships without the aid cf a scale. (3.5) ("by eye"), It is a siirple matter of checkirg the consistercy between known elements and concluding that these inccnsistencies which are in the minority are incorrect. This form of testing is given special note here because of its common use in all forms of arhitEctural design and because-it was used, with few exceptions, to estirrate the state of the generating model. TIESTS CF CIRCUlATION A trief examinaticn cf the arrangement in Figure 4a will show the reader that the seccndary criteria of circulation were fairly well solved while those relating to emergency circulation were not. The access between the various sleeping quarters and their lcckers and washroons was quite adequate. But unfortunately, it was impossible to get frcm any of those dorms to the Apparatus Room without encountering considerable congestion and many obstacles. the ladder Crew Dorm (L-DF) Surely the route from through the Lockers, the CHAPIEE THREE: EXPEEIMEET Washrocn, and the Mechanical Boon fails to meet the The same *as true for definition of direct circulation. nearly all of the major spaces to the Apparatus Ecow. As in the tests of the model, most tests of The diagrarr in circulaticn were performed visually. Figure 4t illustrates the raths traced mentally from the various spaces to the A-RE. The guidelines for making the paths were 1) not to go outside the arrangement and 2) to go as directly as possible between destinaticns. otvious. Room is The congestion at the Mechanical And nothing seems sillier than having to pass thrcugh a Tashroon to get fron one uajcr srace to another. CHANGES TO THE MODEL Proximity was specified between each of tIc dorms and the A-HM as an indirect representation of the objectives of direct circulation. specified between the Apicn and Alignment vas also the A-BM. THE TUIRD GENEEATION At this pcint tLe designer felt he had developed a fairly complete Ecdel of circulaticn. quickly showed his oversights. The computer Figure 5a shows an alternative which alaost ccmpletely satisfies the generating model but fails to satisfy the prchleir. TESTS OF THE MCDEL As in the last interaction, the latest state of the CHAP I ER TER EE: E XPZRE E NT specification nmatrix was cciimpared to the arrangement it generated. as virtually solved, except for The model was clear that the model was cver it one anomaly: constrained and inconsistent in Storage and Vechanical the Hose, its logic relative to Rcoms. It was seen that the model had declarEd all three rooms tc be adjacent, all to abut the A-RM, and all to be That is a gecmetric impossitility. rectangular. Cf course the proximity relationships were not intended as absolutes. TIcy bad been specificd primarily to draw not necessarily touching. the three spaccs rear, Hoever, the discovery that this set of relationships his was contradictory caused the designer to re-evaluate translation of the architectural space program. It was concluded that though the Storage Room was a relatively large space, it was in fact a conglomerate of nincr spaces which were mcre apprcpriately related to cther functions. Because cf the gross scale of this model,.it could easily be eliminated from the specifications and assumed as dispersed in other rooms. Morcover, there seemed to be no persuasive argument for having the Mechanical Fccm near the Hcse- Tcwer as opposed to near any cther space. such an argument, it seemed And in the absence of unnecessary to influence arrancement with such a specification. between the Hose and Iechanical the So the proximity rocms was eliiTinated CH AF IE R THEEE: EXPI JE NT along with the Storage. TESTS OF CIRCULATION The circulation was clearly better than in previous schemes. from the the Direct access was indeed rossible rajor activity areas cf the building. tracing the paths of hypothetical firemen, it But by uas clear that the chjective fcrbidding cross circulation had failed. Pity the member of the engine crew who, taking an extra moment to grah a cooky, is in run over as he darts from the Dining area to the far side of the A-RH. And it is rot toc Lard to imagine a Buster Keaton as the Engine Officer being trampled in the exit of the Ladder Porir Cops. What has happened is fairly ccmmon; concentrating cn cne rrchler, Ly the designer failed to consider the contingency cf cthers. The model had been modified to avcid tcrturous circulation, only to be catisfied with severe cross circulation. ADDITION1AL SPECIFICAUICES A satisfactory representation for all the objectives of good circulaticn had to be developed. was hypothesized that if It the different crews' quarters were on opposite sides of the A-RM, and if the respective officers' quarters were near their own crews, then the two groups would never have to cross each others raths in an eTrergercy. This concept also CHBAPTER TBEEE: 57 EXPIRIEENT necessitated the location of the Recreation and Kitchen-Dining areas on axis with the long dimension of the A-RM1, so that men from both crews could get to their trucks Uithout crossing one another's paths. This was tested by a mental check to see if such a specification could result in a satisfactory solution to the circulation. A diagraratic representation of that test Since no paths crossed, can be seen in Figure 6. there was at least a possibility that the additional specifications would solve the protlem. CHANGES TC THE MODEL As already uentioned, Storage was elimi-nated and the Mechatical Rocm - Hcse Besides this, lower affinity was broken. the designer dccided to push the Hose nearer the street where be could use it to create a "more interesting" facade. The new specifications for improving circulation were also added, and the model was run again. HiE FCUEIT SERIES CF AFERAGIMENTS Despite the pretest cf the circulation criteria, arrangement could be found generating model in its vhich wculd satisfy the latest form. Apparently it was impossible to satisfy all the proximity relaticnships while maintaining the spaces in and procrtion. no their appropriate size CHAP'IER TEREE: EXPERIIIENI An examnination of Figure 7 will show that whereas the majority of the proximities are nearly satisfied, certain significant gaps do exist. horeover, uany cf the spaces are unreascrably propcrtioned. Apparently an unresonatle ccnflict had been created by the number of spaces constrained to occur within the area ringed by the locker and the two dorms. The E-OF and I-CF could have moved outside the ring except for their futual affinity to the W-OF. The same was true for the WASH and SHWR. TESTS OF TlHE MODEL As in the previcus interactions the model matrix time. However, and pushed, it one specification at a satisfacticn, was checked for since so rany spaces were being squeezed was impossible to be certain of many of the visual tests. Tor examFle, it was not clear whether E-OF was undersized in the arrangement or if it was capable cf changing prcporticn in crder to meet the Washrocn. Morcover, error vas. it was imicssible tc know how severe the For example, it was obvious that the Becreation Room was undersized, small it really was. but it was not clear how This information could have been determined by using a scale which matched the display tc measurc tle various spaces in unfortunately, guestion. the designei had neglected Put to display a CHAPTER TER EX1E: EXPEE,LENT constant scale symbol with which the measure ments would have been inpreved. This iculd still have teen an arduous process and somrewat inaccurate. information desired uas primarily Since the guantitative, and since the computer already possessed that data, it seemed much more useful to develop a routine which wiould use the actual data cn which the ccmputer operated to conduct thc tests. The actual operation cf the test routine is described in the next chapter. It enabled the designer to test all the specifications on a space, to find cut if they were satisfied or not, and to know hcw large an error existed. Since the rcutine showed what specification actually existed, the constant rechecking of the generating model was greatly simplified. Using this routine it vas found that all the spaces at the rear of the Apparatus Room were significantly undersized, Dorms, out of proportion, Recreation, and narrow. or both. and Kitchen-Dining The Cfficer's were all tco long Cbviously, all four could not fit within the narrow dimension of the A-BM. FINAL CBANGES TO THE MODEI The objectives of the architcctural prcgram were reviewed and compared to the generating model. Clearly the F-D and EEC had to be near the axis of thc A-PM since men from both crews uculd be using the space and CBAPTEE THEEE: EXEEIENT no other location would picvide the equal direct egress necessary. the two officers rooms On the other hand, were related only by proximities to the sare space, which was now Ladly stretched. h-CF, resolve the ccngesticn In an effort to at the rear of the fire house, it was decided to split the W-CF into two equal spaces, for each of the officers. one The proximity between the old W-OF and WASH was not respecified between the two new rooms. THE FINAL ARRANGEMENI The arrangement in Yigure 8 shows a generation which ccwpletely satisfies this model. computer Tests with the routine shoued that those specifications which were not satisfied had cnly minor errors and were well tclerance within the acceptable cf the designer. Since the ma-jor part of the specification was met, it was decided tc look for anomolies outside of the specification. CTHEF CEJECTIVES The vcid Lehind the Kitchen-Dining area was curious and caused questions "Could it remain?"; such as "What "Could it could fill it?"; be a courtyard?"; then doesn't the L-:EM have one?". "Why All of which raised the issue of the qualities inside the various living quarters. This Nade it clear that previous schemes had CHAPIEE 11HEEE: EXPEETME11 T 6 61 not been so generous to the Cfficers Quarters, which could notw have exterior views. This led in "Why turn to the question, K-D and EEC have an exterior exposure?1 can't the The idea of flipping the two spaces with the Lccker group was considered. A quick check cf the model showed that the proximities to the A-RM would be violated, but the designer questioned the necessity of those relaticnships in the light of the rew fcssibilities. A check of the architectural rrcgrar showcd no specific requirement of adjacency. The drawing in Figure 9 shows the designers own arrangement of such a scheue. arrangement It is possible in this to get frcm the two rear spaces to the A-RM by going through the Locker Rocm. It could be supposed that this would not he a bad feature since the firemens' boots and coats could be stored there. In fact, if they were stored there, the last computer generated arrangement failed to provide direct paths from the dorms to the lockers to the A-EM. However, the need for such a path was not a known fact - only a supposition. It was equally pcssible that the coats and bccts were kept at the mens' staticns cn the trucks. ADDITIONAL INFORMATICN It was clear, then, that additional infcrmaticn was needed in order to continue, since decisions cculd no CHAPIEE EE: hXPERI;ET IE longer Le lased cn kncwn preferences. Further arrangements would also have to ccntain more detaiid information. For example, knowledge of the placement and size of doors was critical for further testing of circulaticn. And scue ccncept of the surrounding environment was necessary tc evaluate the value of providing the dorms with external exposures. The study was stopped at this point. TEe purpcse of the study had not been the total solution of the problem, but tbe exploraticn of testing. By this pcint many observaticns regardirg that objective could be made. IV. CBSEBVATICNS The most obvious lesson of the study was that the designer tested his Ircblen at many different levels. He not orly tested the generated arrangements for their. satisfaction of the generating model, but for their satisfaction of both the architectural space prograu and his own objectives as vell. lie also tested t1e fit Letwieecn these differcnt descripticns of the problem to determine how wiell the generating model represented the space program and to deternine how completely the space program represented the many aspects of the prctlem. This fcrmed a complex netv;ork of interdependant tests, CHAPTIE TEEE: EXPEFJNMEi T all of which were necessary for the continued improvement of the design. It appeared that the first arrangements wcrc tested cnly against the generatirg so poorly solved, there was mocel. little And since that was value in testing any other asrects of the problem, arKra njeament is bys i!Etested fe~' model The later generations provided arrangements which largely satisfied the criteria specified in the model while failing to meet the ctjectives of the space program. In this case, both the arrangement and the model were tested against the larger set of criteria of the space progiau. CHAFEPR ILTEE: And in EXPEFIVINT the final computer generated arrangement, the form met the specificaticns of the model and the space program but failed to satisfy the designer's image of the prcblem. mode5 61C ce - ro 10n1 This is not to infer that a series of tEsts will always proceed so clearly. The differentiation ncted in this experiment was largely due to the computer aided nature of the process. Mcreover, as the model and space prcgram the tests of the model, aere satisfied, program, and mental image made an untraceable series cf jumps, and only the major catagorics or levels could te identifiEd afterwords. It dces seem reasonable to conclude, however, that tests at these different levels are common to all design processes. KINDS Of IESTS: STATE OF TEE MODEL By and large, tests of the state of the model were made on the basis of visual inspection and ccnparison. Since th~e specificaticns cf the model were all geometric CHAPIEF THFEE: EXPERlEENT relationships, they lent theuselves to visual analysis. Only when it was clear that large sections of the model were out of order was it necessary to use the computer testing rcutine. The visual tests consisted of a set of ccuparisons between kncwn conditicls in which the space that did not appear to meet the patter was concluded to be in error. This .uight be characterised as a ccmparison of the actual form with a ngentally Cx PectEd form in which the measurement of each form was made on the basis of further ccmparisons (either nental or visual) one space and several cthers, Letween where the several others were used as the base or "yardstick". M(F) = asau. compwred wrh other The test was passed or failed on the basis of the equality of the results of the two measurements ell - on how the two forms resembled each other. M(F)acbuI M(F)cJeri red CHAPIEF TEREE: EXPELME The computer aided rcutine, by contrast, vas based on the ccmparison of rumeric values determined by algebraic equations which represented the relaticnshir to be satisfied. And since its base of measurement was absolute, a couputer aided test did not need the additional comrarisons with which to establish a point of reference. The two types of test procedures established results which varied primarily it accuracy, but did sc thrcugh ertirely different processes. 0 STATE CF THE ClHEB CEJECqIVES Tests of chjectives cutside the scope of the model uere similar tc those perfcrmed to determine the state of the model. charactErized That is, such tests could all te as costiariscns cf graphic patterns. should he clear from This the diagram and discussion associate:d with Figures 4b and 5b. It shculd be pointed cut, however, that this study was a simple excercise in basic spatial arrangement. It did not get to the level Cf' detail to which mcst of the architcctural rrcgrar had been written. As a result, it was never possible tc test the quality of the flcoring or the location and the number of sill cocks. The only kind of tests which were Loth relevant and pcssible were CHAPER THEE: EXPERJMEVT those of basic spatial relationships - all of which could be tested graphically. While it is obvious that all the specifications in the model and the space program vere exhaustively checked, it is unknown how ruch cf the designer's image was ever tested. And it is just as unclear what caused certain aspects of his image to le tested as oppcsed to others. This type cf infcrmation could not te determined in this experirent. CCMPARISCNS BETWEEN PRCG311E, DESCRIETIONS Satisfaction of criteria at cne level resulted in an escalation of tests to another broader set cf objectives. If the arrangeent satisfied the model, it was checked against the archtectuaral space program. This in itself is a subtle test cf the relevance of one set of cbjectives tc another. comparisons However, more obvious than these were common. Most objectives, such as direct circulation, are open to scme interpretation. By using the ccmputer as a generatcr, the nature of that interpretation vas necessarily very clear; there could be no ambiguity regarding what specification had been used to represent a certain objective. However, this made it possible and necessary to test how well the 1oth CHAPTEF THFEE: EXPEIPIENT interpretation (in represented this case the specificaticn) its objective. lests between different probleml. descriptions were usually made as apriori compariscns of the type of soluticns which were acceptafle to the different elements: the objective and its it reprcsentative specification. that the interpretation If cf an objective could be seen would allow solutions which fell outside of the range acceptable to the objective, it representation, did not pass as a reasonable This ccncept is diagramed in Eigure 10. Circulaticn prcvidcd an excellent example of specifications which failed to adequately represent their objectives. The circulation.objectives were interpreted into several different sets of specifications before a satisfactory model was fcund, During tle third generaticn, the designer considered several rcssitle representations for he could have specified. if circualtion which Each was analysed tc determine an arrangetent might satisfy the specificaticr not the objectives it represented. but It became. clear that no single relationship could adequately guarantee the higher level objective of eiergency circulaticn. Several relationships vere finally selected which it was felt wculd acccuplish the objectives desired. These arc noted in the discussion cf the changes to the model in the third interacticn, and the form of their CHAPTaER RIIPEE: analysis is IMPACT OF T EXPERIH'EN1T illustrated with the diagrao in Figure 6. HE TESTS The effects of the various tests were tcc numerous and varied to draw any general conclusions, Those tests which failed cften lcd to cther kinds of tests concerninC: which specifications had been made; the adequacy of the specification for its objective; the significance of the objective; the existence cf conflicts between differert cbjectives; etc. On the otler hand, successful testing also generated additioral tests regarding: the corpleteness of the model; the satisfaction of the architectural program; the satisfaction of the designer; etc. all lead to different kinds of actions. At scre point, however, cne of the following set of tests is the prccess halts: ansvered negatively, rade and 1) Is the information at hand sufficient to continue? 2) to continue? 3) Is These Are the resources sufficient more to he done? the design is If any cne is arrested. FIGURE 2: MIATRIX OF GENERATING SPECIFICATIONS P= PROXIMITY VA VISUAL ACCESS ALL SPACES DECLARED EXCLUSIVE, AREAS SHOWN BELOW SPACES" NAMES, ABBREVIATION BENEATH NAME SAME AS SHOWN ON ARRANGEMENTS F 7m)F3 AF%.ft!rF FIT FROV FI RTE,-,~ I L- D~M W-OF A-RMA '-RM I STOR ELOCK ~S VWVAI ~ PFRON L t CJ, 7,7 4, F~~1 P,rt, 4 V IQ nF I SC. r!7(Y' Jnrr h iN pF' A T ItO n FI"FT !Z!\ 10 t I C-!-L,,V FI . UE I 5a: 'IT 5n:: F /1'r2/! f Ft CURE FrOM THI OF ( ICUL/\T C ! GEI'ER)ATIO1" r EIREf 7 . FIGURE 6: DIAGRAM OF DESIRED CIRCULATION 75 F I ('U!.~ I ~'L C',- UTF r)!, 1rFPT F EILJMC W LO -F Aff ' r 77 F!GURE 9: DESIGNER T S SKETCH OF ALTERNf\TIVE ARRANGEMAENT FIGURE 10: CONCEPTUAL DIAGRAM OF SOLUTION SPACES AREA OF ACCEPTABLE SOLUTIONS DEFINED BY ARCHITECTURAL PROGRAM AREA OF ACCEPTABLE SOLUTIONS DEFINED BY ARCHITECT'S IMAGE OF HIS PROBLEM AREA OF SOLUTIONS DEFINED BY MODEL WHI CH DOES NOT ADEQUATELY REPRESENT OBJECTIVES OF THE PROBLEM AREA OF SOLUTIONS DEFINED BY MODEL WHICH IS AN ADAQUATE REPRESENTATION OF THE OBJECTIVES OF THE PROBLEM Four Chapter A COMPIEF TFSTING POUTINF This chapter describes a computer routine arrangements a designer in testinq spatial assists generated which by the IMAGE computer system (See footnote The chapter discusses the reasons for developing 111) the routine, output how its algorithm is structured, what produced by the icutine, is and how its algorithm might be improved, I OF THE PIRPCSE THE ROUTINE The testing routine was developed to meet needs in which arose during the design experiment discussed Chapter Three chapter needs pointed out in involved the designer's desire accurate tests used Two of the to perform on the set of specifications by I[AGE to generate he had specified how well those specifications It In he wanted to know loth what specifications were being operated upon particular which were spatial arrangements, testing the generatinq model, constraints that what (ie properly or imporperly) -and were satisfied by a arrangement. was possible to obtain this informaticn without CIAPTE R FCU: any CO1PUTEF TESTS the specifications could. many of since computer aid, and the number and be estimated by visual irspectiori type of specifications could be examined by naking a special search of the 1MAGE data structure. the visual tests However, were not uniformly reliable, was difficult search of the data structure and the ard time conISu ming A third factor, design cxperiment, operating on, result of tc know what information and what. decisions it Since that it was the algorithm was was making as a the information II. THIE TEST The test ALGOIPTHM a subroutine routine is called from can access the same data It TMAGE's main program structure algorithm. undergoing development, was still advantageous the had to do with analysis and of TMAGE's generating developrent algorithm wLich vas not significant in upon which TIMAGI's generating routinc operates This data concerning the current state of .the .spaces and the structuro contains the information relationships specified is organized, Letween those spaces. as shown in Figure semi-lattice or directed graph The designer actuates a command to the main IMAGE 11, in The data the form of a (4,1) the test program procedure which calls by typing the test CHAPTER FOUR: subroutine in Figure (4 12, COMFPUTER TESTS 2) A flew chart of the algorithnm is and the P11 program is included in shown the for those who would wish to follow the actual Appendix commands asks the designer to type testing program first The the name of thE space he made of all wishes to test. relationships sp-ecified by the designer. If the name is Tests will to the space named typed incorrectly or if the designer names a space which does not exist, computer will respond exists ", If with the message "No and will ask, "Aire the designer types "yes", The perfcrmnice of the f u'rther tests from the base point, locating the nancd test then locates a relationship which is specified diagram in Figure 11, (also called to that space, each of if the designer has typed the name by the tranches of the semi-lattice from the tase point to the space level of the structure, would then trace a tranch frcm the APRON constraint a From the the routine would search the data structure "Apron", tracing are automatic The routine searches the data structure constraint) such space necessary?" actual. tests point. It the he can try again from this space be level, finding related to constraint, is "node" It to the PEOXR . The other space which found by tracing the constraints other branch back to the space level Cf the structure In the exarple, the A-RM1 would be thc CHAPTE? FCUR: related space. COTIPUTIR TFSTS The type of the constraint and the name of the related space are then typed out. for the designer. testing routine can access the same subroutines The IMAGI uses a set as the rain IMAGE prcgram, subroutines which suqgest chanqes to be made to the Each of these subroutines arrangement type of constraint which can maintains a general two spaces related condition it. if determines it by the specification they fail, If which the specification by IMACE, When called represents represents one be specified to a syace. Each of these constraint sutroutines description cf the ccnditicr of satisf) the that determnes a change which would cause them to meet the condition (4, 3) The testinq routine calls the constraint sulroutine which matches to the test call the specification space. the subroutine In in has found ccnnected the example, the routine would for prcximity checks the arrangement deterriine if it This subroutine of the AP'RCN and A-FM to fact the two rooms are proxiwate, they meet that condition, satisfied a "marker" is calculated which wculd mcdify the test achieve the conditien is The "'marker" specification changes space in are order to and the suggested chanqes are then passed back to the test the constraint If set that the If they do not, constraint If the is routine satisfied, that CHAPTE is if FCUR: COMPUTEF thc "marker" satisfied, shows that the subroutine has been the test "acceptable," TESTS routine types out the message Otherwise, the condition is the significance with which and displayed as calculated violated is the "errcr" The changes suggested by the constraint subroutine are also displayed on a screen adjacent to the arranqem ent of the From this display, the new configuration space can be compared condition of this appears in as it display The test constraints example, is visually with its the arrangement given in Figure existing An exanple 13, routine continues to search for additional specified to the test space, In the would locate ALIGN and VISACC by tracing out it the other branches connected to the test space, APPON Once all constraints specified to the space have been tested and the results displayed, further tests asked "Are necessary?" the designer If the answer is negative, the designer is returned to IMAGE's main proqram If he answers "yes", however, is he can continue testing his generating model TII The specified TFE OUTPUT designer receives a to the test space; list of constraints the current state of those CHAPTPR iOUR': constraints, relative COCPUT rf T.T1 S whetlher they are satisfied measure of the significance a display constraint c( a of their crrcr; and how the space would he changed by the subroutines This information can satisfy the or not; desicner might wish to ask., a number of questions It enables him to rake a quick check of what specifica tions have been made properly Second, he can get a very accurate of how well a condition is being satisfied Father than relying upon a series of visual comparisons, learn exactly if a geometric condition conpletCly satisfied tells cMerecver, hir how significantly The error is This is relatlive, he can has been the error statement the ccAdition is hovever, measurement violated. and could be improved discussed later Third, the display of the changes suggested by the constraint show the designer, how the specification could be satisfied his goes beyond simple testing,- but can te useful to the design process, the designer might. see that a space is specifications, constraint violating several but. that the changes suggestcd could solve himself and retesting hypothesis For example, ther all, fcr one By changing the space the model, he can verify his In this way, he uses the test routine to make limited experi mental changes, without riskinq the unknown ccnsequences of using TIAGE's generating CHAPTER FOTIR: COMPJRi TEST He can choose to control the charges a1gorith m himself, indirect guidance by the IMAGE substantially to be made under his them or allcw directly This systenm broadens TIMAGE 's which capailities, were previously limited to automated generation Fourth, tie display of the suggested changes the developers of enables IIAGE to know both hciv well and how well the constraint subroutines are functioning, the generating algorithrv is sugqestion incorportating this benefit However, is their not directly relevant to this study TV. POSSIBE I-PRCVEETN TS The First, test routine can be improved terms of a space, terrs cf the permit him to ask for This would elimina te tests Alternati vely, of a constraint developing of the other two Figure 12, he could specify - In type onl. PPOX2 would be tested need tests of only the prcxilmity constraints on a given constraints on the APPCN in Those but also in This would type of constraint space several ways. the designer could be allowcd to request not only in tests in his test this case, in terms both PEOX1 and This could be useful in new features for the generating algorithm, types of ccnstiaints more emphasis Future which are always violated may improvements to IMAGE may CHAPTER FOIRI: COMLIP UER1ESIS enable the designer constraints generaticn to temporarily other than a certain routine The ability suspend all type from the to test only a certain kind of constraint wculd he very helpful in sit this uaticn of course, constraint both the pair of spaces and the type could be specified together. firther limit the set of possible tests, This would (See footnote 3.6) ADJUSTABIE TCLERANCE At present the ccnstiaint whether or not a condit icr is magnitude of error is condition is subroutine determines satisfied exactly, the only estimate of how badly the violated Eut since the error upon some characteristic of the test space, of its dimensicns, designer to specify The it is tascd usually one vould be possible to permit the a tolerance, specify that proximity would related spaces were he satisfied uithir 5-% sral-lest dimension For those under this tolcrance, For example, if of the test he could the two space's tests which were failed the error could be displayed either as a percentage of the dimension or as a comparison of two values, the spaces' size and their distance apart ANALYSTS In CF THE FROELEM many cases, the designer tests an arrangement CHAPTER COPiIUEIL FOUP: TE77 S not only to kncw how certain a whole - Tfo address solved and which are not routine could be nonitored, test space in satisficd" ask permit him the problem. greatest error, Naturally the gained frow rest useful to his process with different for the ccnstraint the weighting specifications In this he could easily explcre the significance of value systcns on the state different of the problem's But this goes well beyond the sirple use specification of effect into that area by allowing designer to experiment way, in this borders on evaluation and could easily be extended structures he could type of ccnstraint -This would information to structure or "least Alternately, or even the ways which i.ere in were This routine could to ask for the "worst" for the ccustraint which was in tests and operated by and which cnes were not, permit the designer are the need, this which would determine which spaces another routine satisficd doing as the arrangement out which areas of to find being are hcw well. his problem is but to learn satisfied, specification testing TESTS OUTTSIDE THE GENFT!ItNG At only these constraints prEsent., the generation useful which operate in routine and which have been specified in the qenerating model of would be MODEL , howevor, the problem can be tested. to be able to test criteria It CI[APTE? FC T p: 88 COPIVUTEIR TESTS other than those which are used for generation would he possible, (lower) fcr exaIle, to set up a third level set of specifications structure those in in the data which would contain constraints sirilar the second level generation designer It to but which were not used for (See Figure 14) This would permit the to define a. set cf relationships which were important but which he did not wish to use for genera tion For example, he could specify one set cf constraints for the generation represented its construct neighborhood relationships this functions, another set of relatinships the eveninq influence educational of a school which He could also which represented use of the building - vhich would not a set of be important enough directly the arrangement of double set of relaticnships to the spaccs, With the designer could generate forms for the needs of the academic functions, and them against test the other functions which they were to perform These are the major kinds of improvements additions which can be made to this There are several arcas which the program. These limitations use of and dependance te advantageous lie testing routine beyond the score of are due to the routine's upon the IMAGE system. to permit the designer other than those of qeoretry or Color, It would to test criteria texture, sound CHAPTER YOUR: COIPU ITIR and occupancy quality, factors 'ILE IS loads are but a few of the. which a designe: mijht At the present time, this is IMAGE system will be expanded type, and and additional test ,;ish to specify and test not possible, However, the to include data of this routines will be both possible necessary at that point For the present, the improverents outlined ahovc are the major changes possible FI GURE I I : DI AGRAM OF I W\GE DATA STRUCTURE BASE POINT SPACES GENERATING CONSTRA I NTS VI SA CC' AL I GN PROXI PROX2 FIGURE 12: FLOW CHART OF TESTING ALGORITHM MAIN IMAGE PROGRAM "TYPE NAME OF SPACE TO BE TESTED" r LOCATE SPACE IN DATA STRUCTURE IS TEST SPACE FOUND? r - 4"SPACE DOES NOT EXI ST" \J/Y LOCATE CONSTRAINT SPEIFIED TO TEST SPACE LOCATE OTHER SPACE RELATED BY CONSTRA INT TYPE OUT: KIND OF CONSTRAINT & NAME OF SPACE CALL CONSTRAINT SUBROUTINE \/ i /Z ) IS CONSTRAINT SATISFIED? N/ N CALCULATE AND TYPE OUT "ERROR" I DI SPLAY SUGGE STED CHANGE DO OTHER CONSTRAINTS EXIST?> AN ARE FURTHER TESTS NECESSARY? RETURN TO I MNIAGE TYPE OUT "ACCEPTABLE" Arrangement Displayed on Another Screen TYPE NAME OF SPACE TO BE TESTED aprn 1.00000E+02 0.0000F+00 4t.00000E+01 4,.00 OOE+01 APRN2 PROXIMITY TO 1 A-RM2 0 IS ACCEPTABLE. PROXIMITY Tn 1 Q P-RM2 IS ACCEPTABLE. ALIGNMENT WITH 1A-RM2 0 THE ERROR= -P ,52000E+nO IS UNACCEPTABLE. Q 1.00000E+02-4.2f600E+0 0 t.00000E+01 4.00000E+01 A-RM2 ARE FURTHER TESTS NECESSARY no FIGURE 13: TYPICAL DISPLAYS OF TEST INFORMATION Fl GUR7E 14: DIAGRAM4 OF POSSI BLE DATA STRUCTUR-E BASE POINT APR P-mA SPACES GENERATI NG CONSTR~A INTS NON GENERATING CONSTRA I NTS PROPORT ION PPN DIX7 I A2C1ITECTURAL SPACE PRtOGRW FOR A FI STATIONC AREAS ATD USES 2,800 Apparatus Room Patrol Room ... W.0C. ........... Engine Dormitory Ladder Doritory .... Firemens 'lash ....... s.f. 120 25 2,945 s.P. Hose Drying To rer..... Hose Store Room ...... Dousing Shover ....... Shower ... Linen Closet ........ Locker Room . 75 75 20 130 Clothes Drying 1oo .. Oil Store............. 30 Generator ........... 1?0 Boiler and Iechanical. 400 25 Air Compressor ....... 30 Janitorial and Clean 905 Fireiens Clerical Supplies ... Janitorial Cleaning . Household Utility ... Fuel Tank ............ Jk. Engrine Officers Room Ladder Officers Ioom 50 . * . . 700 150 150 50 . 650 2,400 s.f. 25 50 50 300 300 250 850 s.f. 150 150 Stairs and Circulation: 700 Total Ab,,ove: 700 125 s.f. Officers Tiash ....... Diesel Pump.......... Fuel Tank .......... . 4,550 Kitchen Dining ... IRecreation .... Telephone Area ...... Total Above: 320 430 10 810 s.f. 4,185 s.f. Fire Station 1 APlAiTUS 20l0U, 2 door, for Engine 32 Each bay to be minimum and Ladder 9. dimensions 70' x 20' as of 1,400 s.f., 's 1 and per Firehouse layout criteria ... ... ....... 2 ...... ... ... ... ... 2, PATROL I002, to contain control console, with seating provision for duty fireman, and commandin; functional vie.,s of the apar atus room and the firhouse apron .... and fronting street . LAVATORY and U.C., 00 square feet 120 s.f. for use by (1)- The duty Patrol Room fireman, ( ) The public, To be located personnel on ground floor. adjacent to Patrol Room and main entrance to enable use by the duty patrol fireman and use by the public without undue station 25 s.f. penetration............................. The apparatus room will have particular proii vision for heating, washing don, air hose, water hose, door control. 1 HOSE DRYINGr TOiER, camable of dryinj hoses ith one ,s fold only. of fifty feet leng of the to:er to be interior .alls The loer lined with a materi.al proveating damage to the floor of the tower metal hose fittings, The tower will be to have srater drains. so positioned in relation to the pumper enthe handling of hoses gine as to facilitate ............. from the pumper to the to.er. 75 s.f. STORE 00, with a minimum capacity of 1 HOSi Hoses are 75 hoses when rollerd and stored. clesign pipe to be held in troughs of metal and construction ...... ........... DOUSING SHO E. (with deluge valve) to hamve two sho;;r heads and a two per-son capability. Maybe two stalls to open directly onto apparatus room; but does not have to be situated with a particular high speed access or convenience !?ater rust be heated circulation reeuirement. 75 s.f. 20 s.f. 1 1 CLOTiJ-S DBYIG 100, for the specC drying of wet firemen's clothing in quantity. Provide for their personnel of three companies plus 25'. Special consideration to be given to the design of clothing racks, whUich may be of metal tube construction. OIL STORE, 180 s.f. to contain small quantities of oil in lockable ........................... metal drums, 30s.f. 1 GENERATO1 RtOC, containiIg a generator with capacity to carry the full electric load of the fire station in emergency. Generator to have automatic spark, operate on illuminatinr gas, have provision for automatic testing. Generator to be positioned to permit walk around and to have designed air intake and exhaust........ 120 s.f. 1 BOIL:R ' "00U. anmmechanical, to contain heating and ventilating equipment and air compressor .......... 1 Provision for one electric air compressor providi.ng copressed air to engine and ladder. Air hose should be ceitrally positioned and service 1 1 should be from cabove ceiling 25 s.f. .............. Janitorial and cleaninag store, with receptor of 1 3, height, shelves and provision for cleaning raterials and implements...................30 s.f. DO IMITCOY for ENGIT' CG1Cm2Y 33 to contain 10 beds of ample size with partitions separating eaci bed ......................... 1 DORMITCJY for LDr 1 FIiW'N'S Tf~I L00, containing 3 U.C. 's of ample dimensions and 3 urinals, separate from shower room 1 400 s.f. . COLiY 15 as above .......... ............................... FI11EN'S 1103E iT00h, containing 4 showers and a drying bench, also containing 6 wash basins equipped with large, good quality mirrors, ample shelving for toilet articles, good lighting and strong tow-el rails .............................. Note that total and generous use of the shover room takes ploe frequently, andthat space and abundant hot water are prime requirements, plus cxhau:;t system. s.f. .700 700 s.f. 150 s.f. 150 s.f. 1 LIN 1 Stationary and clerical SUPPLI>S STO.'Z... 1 LOCKER 2001, war , 'ell-lit and ventilated; containing I ground of 25 hanging lockers for Engine 33 firemen; 1 group of 25 hanging lockers for Ladder 15 firemen; and one set of 80 bootracks each large enough to hold a "hitch". N CL....... . ... .................... The bootracks are not to be more than 3 tiers 50 . s.f. 25 s.f. high. The lockers are not to be more Then 24" wide and 24"t deep, and to have dust collection prevention devices on top. The lockers are not to be less than 24"1 wide and 24" (leep, and to have dust collection prevention bvices on top. There shall be stout, lockers...... anchored bencheslntaeen the ........................................ 650 s.f. 1 A JANITOALw CL 1 HOGUS~ILD UTILIY 1 OFFUICEIR'S 200O0t for I CIILFY, to contain 2 beds, 4 built-in lockable closets, 1 desk, 1 desk-heigjht, 4 draaer legal size file cabinet, 1 swivel desk chair....... 300 s.f. OFFICER'1S ROOM for LADDER COMfPAY. (Provision as for Engine Comany)................... 300 s.f. 1 1 oFFIc2'S basins, 1 1 JING CLO LT with 18"? high receptor CLEiiIC CLST..................... :AS and SLO ER ROO containing 2 wash '.C., 1 urinal, and 1 shower with 2 head ...... ... ... .... .... ... ... .... ... ... 50 s.f. 50 s.f. .200 s.f. KITCH-D1ININGR00 with provision for gas cooking, and a waroing even, for the storage of dry foods and food under refrigeration, and for cutlery, ccockery, etc. Dining will be off 2 tables, each about 4' x 6' and the room will be so proportioned that the tables will normally be used end to end, ie.4' x 12' ........................................... 320 s.f. l adjacent io thec LkitchenICRETIONi 200c, diningroora, and divided fro- it by a flexible curtain pall to thus cnabling joint use of .---..... .... ..... ..... thlese rcooras0 ..... -... 480 s f. There should be provi sion in or near the above roorms for a cigrarette iachine, and 2 slot machines dispensing soft dru'nks and candy. In this living-dining-rocreational area there will be provisional for a wall pay telephone for use by station personnel with a miodicu of privacy. any) and rising to second floor (if Allow for stair Al11ow; for circulato top of hose drying to;er. ... ... ................................. tion. 10 s.f. 700 s.f Fire House Criteria (5) ODSERVATIONS (1) A two story building lends itself better to rapid alarm-generated circulation than a single story building. (2) When firemen respond to an alarm sinal sounding within the firehouse and response is a reflex action - and partially blind and irmpetuous. For this reason the cnergency routes to the fire vehicles must be ms direct as possible and free from obstructions and blind spots. (3) There'will be ample, clear, circulation on the anpoaratus floor. Equipment will be wall or ceiling hung, walls will be vaterproofed as will attendant equipnent, the entire capable of hosing down. (4) Drains in the appiaratus room will be ample, contain traps, and be of sufficient strength to support the vehicles. (5) There must be a heating system in the apparatus roora capable of de-icin-; vehicles and equipment quickly. (6) The apron in front of the house will have a minimum depth of 30 feet and will fall away from the firhouse. Cliectric snow relter mats shall be used with an automatic control to prevent icing of rap. (7) Vehicle Apparatus 2oom ')oors -!ill have a height offinmum 13 feet and minimum of 13 feet. Doors uill have an automatic opening system, but will be cpable of manual operation in energency. (8) Bumpers .ill be incorporatced as apparatus doors to minimize contact damage. Such bumpers shall ba at ground level. (9) Patrol Room should be in optimin command position, vie wing Apron, street and Apparatus Room. (10) There will be a flagpole. (11) Landscape and site ireatment will be such as to reqluire a mininmun of upkeep. (12) The Gasoline Pump will be crd operated and lockable, enclosed for security--yet naturallywented. and 100 OBC VATI; S (Cont' d) (13) A changeover from gasoline to diesel fuel is anticipated. tanAi to diesel Allow for the future conversion of one full oil, and the installation of a diesel fuel punp. (14) The Patrol Room rust enjoy clear functional views of the apparatus rooi., the apron, and the immediate highway. (15) .C. on the ground. floor is for use by duty personnel The The public may also use this facility. in the patrol room. It should be DositioneC to -iinimize the consecuences of temporary absence of patrol roorm fireman from the patrol console, and also to minimize nonetration of firehouse premises by the public. (16) Eercury switches with stainless steel wall plates have Circuit breakers -are proved cfective and pleasant. preferable to fuses. (17) Intercom and public address systems throughout the firehouse. (13) Eledt'ric sockets in proofed. (10) Fire Fighting can be a dirty business for men and machines, In such arears as is and make cleaning an absoring task. practical, surfaces shall be washable and capable of hosing This may consequently rerquire floor drains and srlo down. to floors in domestic as well as firefighting areas. be 0illprovided apparatus room areas shall e w-ater- fes (20) Firemen' s U.C.' s and ;ash basins shall he separate from showers. (21) dooms are to be provided with receptLors, the Janitorial are -hichto be bet-:een S" to 107 high front walls of (22) Good quality locks are recuired throughout andthere will be double locking systems on personal lockers. ( Personal lockers 33) (24) The walls of the base portion of the hose drying tower must be surfaced with a cushioning naterial to prevent at hose ends. damage to metal fittings (25) '1acks for hoses are to be coustructed of metal pipe of Care will be taken to avoid sharp round cross section. surfaces or corners on racks which would damage hoses. ill have dust-free tops. 101 OBSVxTI ONS (Cont' d) (26) Imaginative use of color is firehouse.- (27) There will be a heat actuated alarm system: installed in the locker room, linen closet 'and boiler room. (3) Ceilin% hung light fixturcs are preferred, fitting where prac tical and ossible. (29) It expect throu2out tie as are flush is comimon practice fore firehouse personnel to maintlehone at privauo ese. It is commonly a pay station phone. Provision will be imade for this phone, and directories, and a miniimal writingarea or shelf. tatin one private (30) All doors on -round floor to be lackable from the outside, (:ith self-locking latch locks.) but with inside manual opening cacCpability. (31) There will be a separate rater sup)ly line for refilling booster tanks on apparatus on ground floor. Valve and water line shall b l-V inc'L diameter equipped with a 1-1- inch National Standard Fire Depart::ent male thread gated outlet (32) Ample sillcocks are ar propOrly located. (33) Doors on circulation routes should have observation windows of shatter proof glass, and should swving, as eat asset to any fire station if appropriate. (34) There will be raised thresholds for the containment of water in hose dorn areas, where practical. (35) The consumption of hot food nay be delayed by an alarm, c onsecuently a varming oven i.s required in the kitchen. (36) Ihile there must be a forced heating device in the hose drying tower, care must be taken to prevent excessive heat capable of da0ainS hoses. (37) Iose (38) Dormitory beds shall be at least 6'6" lon; and 40" wide. to.;er shall be veiLed to prevent condensation. 10 2 OBlV~AIONS (39) (Cont' d) mas.culine crcu:staces, Firemen, in res;idence undier tot naked or halfmCay be observed through firehouse ;'idows The fenestration desigrn Ihall dressed from time to time. minllimize tlis visual po7sibilit y. an.e should be (40) Kitchen cooker shall be eledtrical. connercial heaxvy duty tvpe. (41) All enterior nedestrian doors to be self-closing. (42) The temperature control device for the building heatin'30 automatic, but capable of coolin-g systen slioul The nature and position f"icer. adjustment by the Fir should be such as to prevent. o1 the adju;tment con'trl indijscriminate use by firehouse personnel. (43) All doestic, 4-) All polos office, shall and hall areas to be air conditioned. be shutterd. Item 45 the nis-Simple becauoe the human numbers are sCl, that d7onestic take should not be nade of assuin' kitchen appliances are suitable for fireh:ou;e use. it ha- been de nons-trated thaT t si:1 For exampl, A.ll to installed in the pst .which were too have bec Also, the conveniently reccive firehou e fryin-g pan. eitchen efri erator rovec to be too ur;uJl dom stic ood smiall for stations in ;;hich there is co unity purchases and stock piling and the daily storing of lunches carrieC in by the firemen. Iten 4.6 Fireen Flooring in a firehouse m-ust be heavy duty. have to not Floors should quite comronly wear boots. be :axed as firemen in a hurry, novinS quickly, and dres;sing as they ravel are inatInger of slippng. Additionally, extremecly light colored flooring in a : firehouse is i Iten 47 suggestions concerning the There have beei previou use of color and the creation of an attractivo fireIo-re:er, it has been observed in house environnent. at least Pro instances that architects in the past have used attracLive, heavy duty, -;ater-proof wall parers in the domesTic rfos;, but have made the mistake of haxvin collectors, te:;tured anidthey have become dirt the:e paprs The architectsm not easily ashable and requiCinv crubbing. this well-intentioned mistake. are cautioned ai-nst 103 Item 48 of the dormitory shoul d be provided with The -aIls bed lights so that individual Fireen miay rea anight awithout distubing the dor itory. Item 49 Full use should be rmde of the art of acoustics provide for maximum quiet in the domitorics. Item 50 Firemen have to familiarize themselves with the geography of the station' s immediate territory. Additionally, they must be familiar with adjacent on their firehouse territory, which they may visit it this reason, For clergencies. in owi ap aratus more or three for is usual to provide wall soace large maps readily and frequently -ecA by firemen near the Patrol Room is A po.,ion in passing. preferred. Drin'ing fountains should provide cold, water. Item 51 The architect should pay particular attention w;hen designinr the Patrol Room to running- card index or A typical station may have fcom assignment board. 150 to .00 items listed and occasionally up to a The possibility of a rapid usie maximum of 300. or a display under a glass index, circular file, Note that been suggested. have table topped control and con'er uent change to subject are items individual resoe of worthy Tais is removal and replcement. Immediate design. search and an o-pportunity for and readability is a pri-e requireent. visibility Item 52 Before conuencing a firehouse design the architect should spend one full day in a firehouse with Andre Department. Anderson-Beoll of the Public Facilities Item 53 It is not enough merely to provide automobile parkAutomobiles may be subject to theft ing facilities. security precautions are re-uired. and or vandalism Item 54 required there are Boecause of the high visibility often large areas of glass in Patrol Room, and just As a as often, no special provision for insulation. result, Patrol Rooms often become exceedingly hot or The architects are cautioned against exceedingly cold. this mistake. Item 55 'ahen an alarm sounds the Apparatus Room doors are openedi and the entire firehouse [,aff denarts the The firehouse doors close firehouse on the vehicles. lights go out autofirehouse the automatically and leaving the firehouse minutes, several matically after freThis situation and its unguarded and in Jarkness. quency should be noted and provided for, particularly in terms of security. to 104l Ite. 5 Uhen an al;rm is sounded at night,house is il uaated. the darkened fire- Item 57 The impact of the alarm on a sleepinr min can have side effdcts. The Firehouse internal illumination and the alarm bell system should be on separate s-itches in the Patrol toon so that lights rmay be siitched on one or tw:o seconds before the alaru is sounded. Item 58 There shall be a provision for 1ow-level lighting in thc Anparatus 1oom at nigAht when the main illumination may i.ormally be sitched off. Item 59 It is not consiider-Vd an advantage to have a splendid, smooth surface finish under -et conditions. A floor with too rou-h a surface ill hold water by "pocketing" and canot be satisfactorily squeeee'd. Item 60 Incredible as it seem, Cay many instances have been observed in firehouses in Which doors did not swing in the direction most apsropriate to facilitate rapid movement by fireen responTingv to an alarm. Also observed were doors hug on the sidc of the door frame best suitable to impeCe novement under these cir cuL tances. Tie designers are ;:arned against this thoughtlessness. Item 61 In keeping with the principles of maintenance and cleaning , rsee roller blinds are preferred to venetian blLds-, in a irehouse. The designs should be constantly :ara that there scarcity of tir-m for cleaning and maiance throuhout the fi re house. The architect shoulc enacavor to esign dust-free kitchens andto incor)orate surfaces nhich lend themselves to clean by firemen. 105 Apparatus Room :iging vehicl (* 40' min ~ 50' miIn 90 feet minimum The apparatus should have two thirds of its length clear of- the apparatus room before commencing the turn, i.e. minimum 40'. A turning radius, measured from the outside of the vehicle commencing the turn, of at least 50' shall be allowed. 10C Maximum Apparatus 8'0"w by 62'7" long The design and layout of the Apparatus Room must be based upon this maximumvi vehicular unit although in practice the space may be occupied by one or more shorter units. A width for circulation is allowed between the apparatus and the long wall, minimum of hence the outer bay width 2 0'. 70 feet length minimum -- y- 30' min APRON II may be advan-tageous to have one rear entrance if site conditions a I i ow. ExitEntranc 13' w min I 13' j min 107 APPENEJX FL1 II PC-PAV FC1 COMEUTE TESTEF: FFCC IIEIF T1STING (IErWGNC, V,SChL F).; DEC LARAlI CNS DCI PET(L) IAEl., (SXY , IFI RST INITIAL (') ) .TATIC, (IN,P P (250,:3, 6) , T ( ,V ;IS3AS (250),7) ,VP)I :S2 FIX D STATIC, (SVNAME ,II ,SVN,ISV ,JSV, AI,SVV AL,SVCST,SVI) EIN F71 XED STATIC (SVN2, KT 1 , 1X, IY,OTA 3(2 S) S USI,A TA3 (Q) ,SVVC) FIN FI XED STATIC DCL EXFC RETURNS (ECI1NTEF), S ET (3: 2 ) B IN F IXED PASED (20 , 1 UDATA SThqIC, 2 UIA VE CHAF (9) VAFYII"G, 2 VARS(9) 2 UV1 EIT (9) 2 MDCE1 11IN FIXEF, 2 -CDE2 BIE FIXEL, 2 fEIF3, 2 r U 14 GVARS (9) 1 CTF1 (3) STAT1C, 2 DTYPEX EIN FIXE, 2 DP TCIENTEF, 2 EL EIN FlXED, 2 1 BI ,' iX T7Pj (I, S( T~ N , )1 1STIf 1 JN F1 T4 I-IC I t< Nr N INN r X SU I~SHEb) STI AT IC, 41-CI ) , L1 3T FIXIE ST TAT I, I N FIX, INITIA",L () SAEAT ICF 2 CT I) BIN FIXED STATIC (USE INITIAI (1) , CCESTFAINT INITIAL (2) , VALE iNITIAL (3) , PEL INTIIAL (4)) FT XED, INPUT CH AR ( 1) VARYTNG, ID 1G1C CH A R (1) V AFYING, IN 2 Cl-AR (4) VARYING, I13 CHAF (1) VAPYING, ASTR C1AR (8) VARYIN9, 0ST S CHAH (8) VAFYIG, SPACE CHAI(10) VARYING, 13ET (10) lABEl, LAPEL, AP RCC(50) P1CC_ NC(5 I) lAPEl LAE L , RETV (4) CUTCH CH AP (R) ST ATIC, 1 CTAP STATIC, 2 UNA F1 CI AR (8) VAFYI NG, 2 C ITTS(6) T f) (9) , FFI 2 W, 2 CET A, 2 CETA 2, 2 CD TA 3,2 I C1DE, CD TA 4,2 CDTA5,2 CLTA6,r A PPE NDI X II 108 (2 CONCT, 2 SEENL,2 EEC E) BIN FIXED, 2 CNAfME CI!A F (8) , 2 DUP1, 2 DUM2, 1 VDA TA (6) STA TIC, 2 CDAT A CHAF (8) VARYING, 2 FEATA(9) , I 1 (9), 2 M1 PIN FIXED, 2 F22 1IN FIXED, 2 DUV5, 2 DUM6, (V DAT A ((6, 9) , AMAi (1,9) ,C NA T (1,1) ,ANS (1 BIN FIXED, SV !AT A1 (2, 9) ,XM ' (9, 1) IN ITTA L ( (9) (0.9) DECLARE (P,QNE1,F('1) F Ei U RNS (HIT (1) DSTAPT RETRJNS (POI ) 2 UV )) STATIC; IN IT IAILIZ AT ION INX=0; INIT: NEEBIT=2 IIG-=0; IC=0; CLDIDCI=U; INT1=1; 1NT2=9; GET T IEST J=0; CIEEC=(; CLDID=0; INFCFMIATICN DISPLAY ('TYPE NAME CF SFACE TO SPACE = IN21 1 IDWGNO; IGUYID = ID SE (SPACE) ; fE TESTEE') REPLY (IN2); /*---------- TEST POUTINE IS AUTCMATIC FRCV THIS ECINr. SEARCH DAT A STSUCTUPE FOR TEST DATA NCTE: P IS A SPFCIAT IUEFCSE LOOP FUNCTICN TO S EARCH THE EATA STRUCTURE. Q,ID ENT, DSDSPL1, IDSE., AND IDTYEE1 ALSC SFECII EATA S2ARCHi FUNCTIONS. LOCATE TEST SPACE IN DAIA STRUCIURE P1: iF P (D, SUSE)THEN DO WHTILE (Q(1B,USE,N)) ; IDUU=IDEN T (SE"T (USE) ) ; IF (IDENT (SET (USE))-,=lGUYIE) THIN GO TO SKIPIT; Rl: IIG=1 GET ATA ON TEST SPACE IECOM DA'A STFUCTJFE CALL., ESDSPL 1 (SET (USE) ,ADDR (UNA ME) ) ; SVUSE=ST (USE) CIIDC=0 ;J=1 ; APEEN\DifX 11 DO KKJ-=1 TIO 0 XMAT (KK ,1) =VA F S(KFJ); FN E; BYBY:1)C I=1 TC 9 ;IFAN ITJ ID=IDEN I (SET (USE)) DO I=1 7 C 9; GVARS (1) =VA RS (I) ; ENE; ( )=0;ENE; NEWPIC BIANES SCREEN AEJAC FNT TO ARRANGEMINT BEING TESTED IF INX = 0 THEN DO; I N X =1; END; CHIIGDSP DISPLAYS EISPLAY OF CALL NEWPIC(SCALE,0.0,C.0) TEST SppC AS IT ; EXISTS. CALL CIEE CDSP (ITJID,GVARS (1) , CV AES (2) , GV ARS (4) ,GVAR S( 5) , UNAME) ; CL=1 ;AL=1; /*-~-- - --LOCATE CONSTRAINIS CN SPACE IN DATA STTUCTURE IF P(USE,CONSTRAINT) CLDI D=1DENi (SET (USE)) IF OLEC=) THEEN GO N=1; GO IC CPDEJ; RETPT(1) N1; R2: CLDC=1 GET DATA CN TEEN DO WHILE (Q(1B,CCESTEAINT,N)) ;CLIEPT=SET (USE); !, 2; CCNSTFEItT FOM EATA STRUCT CALIL DSDSPI1 (SET (CCNST FAINT) , AEDP (UNAMF 1)); CALL ETYPE1(SET (CONSTRAINT) ,CIRL (1)) I =; LOCATE OTER SPACE RELATED BI CONSTRAINT IF P (CONSTR AINT, VALUE) THEN DO WHI LE I=I+1 ; CALL ETYPE1 (SET (REL) ,CTRL (I+1)) CET DATA ON OTIIER (Q (10 EVALUJEEEL,N)) SPACE CALL ES)SPL1 (SET (VALUE) ,ADDR (CDATA (I)) ) ; IF IDE NT (SET (VAiUE) ) =CL.EID TE EN II=I; ELSE EC; SVID=TDENT (S}T (VALU E)) SVVAL=SET(VALUE) ;SVC S'=E T (CCNSTRAIT) ;Ki=I; END; OUTHEVE: END; END; ELSE DO; IF IC=1 THEN LO; OlEID=0 ; IC=0; END; END; GO TO EONE1; APPENDIX II 110 SnIrIr: END; DONEI1: N= 2; END CF ORDER IS TETl 1UTINEF THE ECUTINE WHICH CALIS T-iiE GO TO CEDEB; RETPT (2): P3: DISPlAY ('A E FUUTER TESTS IF IN3=' Y' THEN GC T1C I IT; RETURN; COlISTRAINT NECESSARY' SUBROUTINES. ) REEIY (IN3); ODER: IF IIG=C THEN D0; D1.S PLAY ('NO SUCH SP 1PCE EXISTS.'); GO 0O BLIET (2); EN£; DO Il=1 TI7O I; DO F=1 TC I; IF M+7'0a=D YPFX (K+1) 'IHEN DO; IF !=II THEN VNAFE=; CO NE=1 TC 9; VDATAU (A, A N) =FDA 'I (K ,N) ; E ND; END; IN E ; ENE; NND EYEX (1) -810; GO TO PROCNO (tN); PROC NC( 1) : C L 01LP, AP( VIA ME,VDA TAU ,C1ITS, ,CDTA, NUEsCIT) DISPLAY (' A T-Cv WI ; f] ') ; CALL FIXLOCO (CE ATA (KKi) ) ; GO TO M'ATSET; SUBROUTlES CALLED ItN FOR OTHER POSIBLE CCNSTRAINTS AEF SI NIL AF WPYS, EUI b.R E NOT SHON HERE. SA T ET: 0 N = 1 TC) 9; A1YAT (J, NN) = VDATAU (1, N\ N) IF APS (AMAT (J, ) )>1.0E '-(95 THEN IRANK (NN)=1; END; C MAT (,1)=-VDATA J(2,1) ; OUT: IF A Z ERO VN!A E IS RPTUPNE 1)l ilE TQEST HAS BEENZ SATISFIED. C9 HEFWISI ITHE ERRR IS CALCULATED AID DISPLAYI I. IF V"A&E=0 THEN DC; TLISYIAY ('IS ACCEPTAEE ); GO TO RETPT (N); END; CA LT. IVMGG (AMAT, XMA'T ,I NT 1 , I NT 2,1 E:Z=ANS (1,1 )-C AT 1 ,1) ; NT1, ANS); ; APP1EDIX II iII T E l ERlP= ' I I E R) ; DISFLA Y ( 'IS UNACCEFT A 1LE. THEN DO0; D0 ICUY=1 TO 6; IF AHAT (1,1GY)- GY) AV A Y =C2 AT (1, 1) /AMA 'I (1 ,IG UCY) ; GVARS (IGCY)=AVApY; END; END;-~ THE CHANC F WHICH WOULT) BE SUB0UTINE IS DISPlAYED. NADE CALI. CllhGDSP(ITJID,GVAPS (1) CDEAIT A (YKKII) E ND ); GO TOPR ETP1 (N) qE S1 IA X ; BY ,GVA-S THE (2) CONSTFAINT ,GVARS (4) ,GVAIRS (5), 112 C S FCCTTE CHAPTIE? ONF From 1968 until the present, the author was (1 ,1) involved in a projec t Vhcse objective was the development of a computer aid for space arrangement That project resulted in a systew of programs which an arrargen-aits to satisfy generated spatial architectural rrobleu spccified by the designer/user The system is called TMAGF "TAGE is a nettcrk of ccmputer programs which generate three dimensi-onal spatial arrarge.mcnts to icnships specified by a spatial r satisfy is ccmrposed of 'pecification A prol: designer two EClements: gccmetrLic descriptions of the spaces to be confiqured; and relationships between those Given spaces which the designer wishes to achieve. a problem description, IMIAGF generates arrangements a confiquration is achieved which satisfies until If no such specified relaticnshi ps. all configuration is possible, IMAGE produces arrangements which mrinimize dissatisfaction of the specified relaticiiships The dimensions, position and orientation cf each space may takbe cn any value in a continuous rance and are not c1onfinod to discrete intervals. The spaces may cvcrlap cr be excl.iisive according to th1e relationships specified by the designer, thereby pcrmitting prcjects to be described in terms of individual rocms or in terms of the and fucticn they serve activities A diverse set of relationships way be used to Thbe designer may constrain the specify problems size, shape, orientation, and position cf any space, and need not limit his specifications to a tine or single type of function, such as trip may be combined Different rela tinships prox imity to represent: specific, complex objectives, allowinq his specification to his a designer to tailor particular method of design It was believed that a computer system which assisted spatial synthesis would enable the designer t c explore varia tiens of his image of a prchlem - to maintain close watch over his objectives - to develcp a wide range of alternative schemes, and thereby, to achieve more satisfactory "i fors tc It was soon realized however, that the alility tectural archi to scluticns form alternative generate problems vas inadaquate for desiqners' needs in at least one ma jor respect: it was incapable of assisting the user in the sirple task of testing alternative FOOT NO TE -S3 113 ,-as in the It arrangements against diffeeint criteria, of mecting this ncd that the study cf testing i:'terest See: was undertaken JohnsonTimothy E , et al, IMAGF: An interactive, Graphics-Pased Computer System for Multi-Ccnstrained Spatial (Cambridge: IITT Dept. of Synthesis Architecture,197 ) (1 2) Seec the discussion of "City Design", "City Designing", and "Functional Analysis of City Design" Porter, in: William" 1, The Developreent cf Discourse: A Languaqe for Computer Assisted City Design, thesis, NIT rept, (Cambridge: unpublished PhD 1969) Studies, Urban of (1 3) An elaborate present ation of the synoptic approach .weaknesses as compared to to problem solving ardthe lasic contcri c,: ,vides pt apprcaches heuristic Lindblom, Char and David Braybrooke, Eccision of A Strategy (New York: Prce Press, 1963) CHAPTER T0O (2. 1) Several well known volumes exist, the lest knowln of which is Ramsey and sleepers, Graphic Standards. (2 2) This Phenomenon w:as observed by Dr, Pohert Weiss during a case study of the Seattle World Fair acne for Other observaticns are reported in: IBM and lBoutourline, Scrge Weiss, Robert S Fairs, Pavilicns, Exhibits, and Their Audiences Report, 1962 Unpublished P CHAPTER THEiEE (3 1) The program was made available for the study by It bad been Ashley, Myer, Smith, Architects, Cambrida.e existing of an remodellinq the for office that by used provided been and had Station, Bay Back house, the fire for the architectural ren'cdelling by the Public ThE Department of the City of Boston Facilities the physical to bring purpose of the remodel ling vas house, built near the turn of the condition of the fire century, up to the quality of recently constructed The program, therefore, was not signi fica ntly stations FOCTNO TE S changed lii frcm thcse used fcr rev ccnstruction (3 2) During visits to the fire house, the architects found that security of possessions was very irportavt to the firemen, Evidently, snccping and pilferaqe were not uncommon among the nen on duty Several men indicated a desire fcr lockahle clcsets and storage areas vhich could be easily and continuously observed and guarded. Consistent to Tiost architectural programs, no mention of needs of this kind was made (3 3) Tc a large extent, the selection of the fire station was made on the ccmmendation by the designers at the architects office that this program was far and away the most complete that had been used in the office. (3 4) Whereas IMAGE attempts to alleviate all errors, at any point during the generation process, a large portion of the rodcl may be in errcr (3 5) A design instructo: at the University of Arizona once failed to attend a student party on the qrcunds that the map to the party had been drawn with "utter disregard for relative distances". He had dcduced frcm other relationships cn the map that the p: . aoit a mile from a locating intersection, wheti in fact it was three miics. Having drive.n the mile and not locating the house, he returned hoie in confusion, (3.6) An excellent discussion of set theory, thc algebra of events, is given in the first chapter of: Drake, Alvin W Fundamentals cf Applied Probability Theory (New York, McGraw Hill, 1967) CHAPTEP FCOR (4 1) For a more detailed discussion of this data structure system see the appendix of: Timothy Johnscn, et al, Op Cit, 197> or Crick, M F lorie, PA. Mosher, E d, and Symonds, A J A Data System for Interactive Applications (Cambridge, IBM Report 320-2 58, 1S71) (4,2) IMAGE operates on an IFM 36,/67 computer using the CP/CiS time sharing systerr The user communicates with the system by a typewriter-like, console, His typed commands, as w:elI as typed and 2 dimensional graphic responses, are displayed on a simall1 optical scrccn, si ilar i to a tclevIsio screeil (4, 3) For a more elaborate discussion of the various constraints possible within IlAGE, the conditions they and the vay in which TPIAGE uses them snek to satisfy, during generation, see Chapter 2: The IMAGE Systcm, in: Timothy Jchnscn, et al, Cp Cit, 197- 116 FIL R1OG RAPU"y Alexander, Christopher Notes on the Synthesis cf Fcrm Cambridge: Harvard Press, 1964 Brayhrocke, David and lindblcm, A Strategy of recision New York: Free Press, 1963 Yastman, Charles Charles Solving Strategies in Design," EDRA1, Proceedings of the 1st. Annual Environmental resign Researcl Association Conference, Henry Sanoff and Sidney Cohn, editors Chapel Hill: School of Design, North Carolina Stzfte University, 197 IProblei? Ju, Perkins, John Weinzapfcl, Guy Johnson, Timothy E Morris, David ScloTova Doris TM1AGE: An Interactive, Graphics-Based Computer SysteM for tulti-Constrained Spatial Synthesis 19T' Cambridge: MIT, Depari-ent cf Architecture, Hall, F I and Schwarts, J.G. Manheim, Marvin L Modelling the Evolutionary Nature of Problemn Solving Cambridgc: HTT, Schccl of Engineering, 1967 Narvin L, Manheim, Problem Solving Process in Planning and Design Cambridge: MIT, Schccl cf Fngineering, 1967 Herbert A, March, James G, and Simon, Crganizaticns New York: Wiley, 1959 Myer, John R, lewis, Krauss, Richard I. Danielson, Scott and Poger Design: A Case History Cambridge: MIT, Center for Building Pesearch, 1968 Negroponte, Nicholas P and Groisser, Leon B. Urban5: An On-line Urban Design Partner Cambridge: IBMI Feport 32C-2r012, 1967 Porter, William I The Develcpment of Disccurse: A Language fcr Ccmputer Assisted City resign UnpublisLed PhD Thesis, MTT, Department of Urban Studies and Planning, 1969 Paiffa, Fcward Decision Analysis Reading: Addiscn-Wesley, 1968 Simon, Hertert A Sciences of the Artificial Cabridge: MTT Press, 1969