A knowledge-based system for house layout selection

Mathematics and Computers in Simulation 66 (2004) 43–54

A knowledge-based system for house layout selection�

Ana González-Uriela,b, Eugenio Roanes-Lozanoa,∗a Departamento de Algebra, Facultad de Educación Universidad Complutense de Madrid,

c Rector Royo Villanova s/n, Madrid E-28040, Spainb Departamento de Ideación Gráfica Arquitectónica, E.T.S. de Arquitectura, Universidad Politécnica de Madrid,

Avenida Juan de Herrera s/n; Madrid E-28040, Spain

Received 27 January 2004; received in revised form 18 February 2004; accepted 23 February 2004

Available online 17 April 2004

Abstract

Automatic reasoning is applied to building design problems with a large number of standardized conditions, suchas normal housing. Industrialized home building can provide quick-to-build low-cost dwellings, but these buildingsoften are not as appropriate for the local climate, building site or to the occupant needs as tailor-made homes. Inthis article we present a knowledge-based system (KBS) that combines advantages from both methods. It considersthe data corresponding to the building site (dimensions, local building codes,. . . ), the climate and the purchaser(budget, family size,. . . ) and recommends a particular design.© 2004 IMACS. Published by Elsevier B.V. All rights reserved.

Keywords:Industrialized building; Dwelling design processes; Knowledge-based systems

1. Introduction

One of the greatest architectural obsessions of the XXth century was industrialized home building.Both in North America and Europe, this cause has had earnest proselytisers; several studies in design andproduction were carried out and there were some encouraging experiences, proving that industrializedbuilding could provide good houses at a good price and in less time than other methods.

But, apart from a number of isolated achievements or some particular public planning policies, indus-trialized housing is having a low impact on large-scale building operations, and people still tend to relate“prefabricated” to “low quality”, considering prefabricated to be rigid and unadaptable, very homoge-neous and to have poor environmental standards.[1,2]. This work tries to show how this situation can

� Extended version of a presentation by the first author at the Young Researchers Session at ACA’2003 (9th Conference onApplications of Computer Algebra, Raleigh, NC, USA, 28–31 July 2003).

∗ Corresponding author. Tel.:+34-91-394-6248; fax:+34-91-394-6248.E-mail addresses:[email protected], [email protected] (E. Roanes-Lozano).

0378-4754/$30.00 © 2004 IMACS. Published by Elsevier B.V. All rights reserved.doi:10.1016/j.matcom.2004.02.020

44 A. Gonzalez-Uriel, E. Roanes-Lozano / Mathematics and Computers in Simulation 66 (2004) 43–54

be improved, giving the average dweller the opportunity to be involved in the design process, with noadditional cost.

In this first proposal, our target subject is the single-family dwellings, meaning detached or semi-detachedhouses.

2. Architectural preliminaries: a historical overview

During the XIXth century, and closely related to the industrialization level the United States hadachieved,balloon framingturned traditional timber frame building techniques into an industry (Fig. 1).The old, expensive and complicated wooden joints were replaced by factory-made nails, so that on siteskilled labor was not necessary any more. The complete house came in pieces, numbered and ready to beassembled. This new system spread quickly and became a standard for detached house construction. Hadnot this type of construction been used, cities as Chicago or San Francisco could not have experiencedthe rapid growth they did in really short periods.

A popular antecedent to modular housing can also be found inmobile homes, from the Wild Westcaravan wagons to nowadays sophisticated recreational vehicles. Now the house left the factory com-pletely built—and on wheels, ready to cross the country or settle down anywhere. Its dimensions,both in width and height (and even in length), were severely constrained by road regulations. Nev-ertheless, e.g. the maximum width grew from 8 to 12 ft and even more in some states (nevertheless12 ft wide models usually require a special permit in order to move from one place to another)[3].

Fig. 1. Balloon framing building (XIXth century).

A. Gonzalez-Uriel, E. Roanes-Lozano / Mathematics and Computers in Simulation 66 (2004) 43–54 45

Later, extensible elements and added modules were developed in order to mitigate these disadvan-tages.

Ideas also borrowed from the car industry, such as considering mechanical equipment as an essentialquestion, or introducing the possibility of having a new article every number of years, were the maincontribution of R.B. Fuller’s proposals (e.g. 1929Dimaxion House).

Many variations on these proposals and some other interesting possibilities were tried—like the 1948 C.Koch’sAcorn House(Fig. 2), which came folded in a 2.45 m×2.75 m×7 m tow and could be completelyerected within 32 h, but prefabricated housing production in the US did not grow significantly until thelate 1940s (after WWII)[4,5].

Most of the first attempts in industrialized housing in Europe took place in the context of socialmovements and urban-planning new ideas. Specially after WWI, many architects (e.g. German Alexan-der Klein) exhaustively studied economic dwelling and minimum home featuring, paying attention toquestions as livability over built area, circulation spaces rate, rooms distribution, natural lighting andventilation[6]. TheOne Room Housedevised by the Russian M. Ginzburg in 1929 contained all the basicliving facilities in a 4 m edge square cell, consisting of a set of factory-produced lightweight elementsassembled with the help of a crane. The idea was to repeat it by the thousand and to sow them on linearurbanisations all over a green city basis.

The French architect Le Corbusier’s 1939 proposal boasted of employing totally typified constructionelements (including mass produced staircases, windows, kitchen equipment,. . . ), all of them assembledwith screws and bolts—dry-assembled.

In the WWII postwar scene, severe economic conditions and serious dwelling shortage led to quiteinteresting housing solutions with minimum material costs. The French engineer J. Prouvé’sMetropoleHouse(1953) had all its structural elements made of 1.5 mm steel plates, bent and screwed to becomebeams and columns.

SECTION SHOWING FOLDING

SECTION

Fig. 2. Acorn House (XXth century).


Fig. 3. Open industrialization (XXIth century).

In the last decades, industrialized housing development has kept making good use of techniques andelements from other industries—mostly car and aircraft industries, with special attention to increasingflexibility and adaptability[7]. For instance R. Rogers and S. Rogers 1969 proposal usedAusterbuseswindows. The 4 cm thick inner partitions of the house were detachable and attached to retractable joints.

But the main goal achieved in recent industrialized building, at least partially, consists on what isusually referred to asopen industrialization. That involves not having to buy all the components of thehouse from the same company, but having the possibility of choosing each part of them from a rangeof different vendors (Fig. 3). A careful standardization of the building elements is necessary in order tomake interchangeability a real possibility. Some of these industrial specification agreements have becomea fact (e.g. 1980 French Association of Construction and Components regulations[8]), but there’s still along way to go[9].

3. Our approach to industrialized housing

We have chosen existing modular industrialized housing elements and we have supposed that open-industrialization is a reality. We have created from those elements a catalogue of possible house layoutschemes that could be employed in an industrialized housing system and that covers a wide range ofrequirements (as suggested, e.g. in[10]).

Prefabricated housing involves standardization of rooms[11]. We have adopted a 3.60 m basic structuralmodule, because it is a proper length or height for both beams and pillars and provides proper domesticroom size. Moreover, it allows us to employ divisors, such as 0.60, 0.90 and 1.20 m, or consider severalcombinations like 2.10 2.40, 4.50 m (all in a 0.30 planning grid) that correspond to the width andlength of usual existing elements, as doors, windows, facade panels, appliances, pieces of furniture,. . . .Obviously, a similar nice measure should be chosen using units in the imperial and US measurementsystems.

In accordance with these modular dimensions, we have considered a range of different sizes and shapesof rooms, which could be assembled in several combinations. These elements correspond to one of the twomain types of spaces we have considered in a house[12]: (main) rooms and service pieces (Fig. 4). Thelatter include kitchens, bathrooms, laundries (with all of their necessary fittings and equipment), storage


Fig. 4. Our approach to industrialized housing: modular service rooms.

areas, staircases, furnace and air-conditioner space. . . . The main rooms or living spaces, not labeled asbedrooms or dining rooms, are intended to be as multiuse as possible.

By combining the different rooms and service pieces—in the appropriate proportion, we have designed184 possible houselayout schemes, grouped in four mainlayout types: rooms aligned in a single line;rooms in a double line; rooms around an inner court (patio); rooms around a service core. The first typeincludes two subtypes:with the service pieces among the roomsandwith the service pieces in a parallelline to the line of rooms.

4. Knowledge organization and description of the system

As mentioned before, ourknowledge baseconsists in a series of criteria to determine which house layoutscheme fits the requirements of each case best. These criteria might be grouped in three categories:

1. Those attending to the climate. This refers to the climatic features of each type. For instance, patiohouses can provide a shadowed area in hot regions, while compact shaped houses would be kept warmeasier in cold climates[13,14].

2. Those related to the building site. They include the site’s real dimensions, average hours of directsunlight, the existence of nearby noise sources or beautiful views and, additionally, the local buildingregulations concerning the plot.

3. The needs of the particular group of occupants, which means the number of people expected to livein the house, whether any of them would work from home, if is there a handicapped person,. . . [15].

The first two are used to choose the main house layout type, while the third identifies the most appropriatelayout schemes, and sometimes—if none of the first layout type fits, may require another layout type.

Besides the best answers, the system will offer the users, when possible, some other proposals fromthe second best layout type, in order to enable them to choose among different possibilities (Fig. 5).

As there is a large amount of data involved, but not much interference between criteria, we have decidedto use a matrix approach to this problem. Prior to building this system, we worked on a smaller system,


Fig. 5. Scheme of the way the system works.

based on building constraints originated from town-planning laws in Madrid, in which we used GröbnerBases[16], but the inference engine turned out to be infraused. For the new system,Maple [17–22]hasbeen chosen because it is a user-friendly system that provides comfortable and complete programmingand matrix-handling features.Autocad[23] has also been used to generate the graphics that are given tothe user as part of the output thatMapleprovides.

The description of the system is as follows: a matrix has been prepared for each one of the main layouttypes we have considered (Fig. 6). Each row of these matrices describes a certain house layout scheme,containing information about features such as number of rooms, area occupied, number of floors,. . . .

For instance, row [e2 1 2 3 6 93 162 15 6.5] stands for a house layout scheme accordingto:

• Layout type: e2 → main type: “rooms in a line”, subtype: “with service pieces in a parallel line”• Number of floors: 1+ 1 = 2• Layout schemes for each floor: 2 → layout schemee2/2 and:3 → layout schemee2/3• Number of possible (main) rooms: 6• Living spaces(usablem2): 93• Total built area size(m2): 162• Length of site surface occupied(m): 15• Width of site surface occupied(m): 6.5• Total occupied area size(m2): 15× 6.5 = 97.5.

Moreover, all the previously mentioned information about climatic passive characteristics of each housetype, building regulations, and proper answers to specific requirements has been translated into (if-then)logical rules and integrity constraints. An integrity constraint translates something like:

the total built area on the site surface cannot exceed the plot ratio indications, what can exclude certainhouse layouts.

On the other hand, one of the rules looks like:

IF usable square meters of living space is above the size needed THEN that scheme could be suitableELSE larger ones must be searched for.

Or like:


Fig. 6. Example of a data matrix: type “rooms in a line”, subtype “with service pieces in a parallel line”.

IF the region is cold THEN do not choose patio as first option.

In addition to that, we have set some other rules to deal with the data introduced by the user (datapreprocessing), such as the one which generates a submatrix containing just rows standing for houseschemes with most of the rooms facing south, once the dimensions of the real building site are known.That is becausein a sunny and cold climate, single-line rooms should be a good choice to obtain directsun warming, but this type should be rejected if it does not fit facing south within the building site. Forinstance, another preprocessing rule establishes the main rooms area needed considering the number ofpeople expected to use the house, added storage space for special objects, if any; room enough for workingfrom home or particular activities that are expected to take place.

Finally, these rules and integrity constraints are integrated intoMapleimperative procedures that checkthe data matrices in order to determine the rows whose features fit best. Note that the whole process canbe algorithmically decided and there is no need for knowledge extraction in the usual sense of ArtificialIntelligence[24,25].


4.1. Example

The user is required to introduce the input data. Suppose they were:

Building site area (m2) 2000Noise sources—direction to dodge (no/yes: N-S or E-W) NoBeautiful views to be considered (no/yes: direction N-S, E-W) NoBuilding rate 0.25Occupation rate 0.10Maximum height permitted (no. of stories) 3

Shape of the area able to be built on:Maximum length E-W (m) 28Maximum length N-S (m) 7

Basic climatic feature Temperate hotNumber of people expected to live in the house 4Number of them working from home 0Number of special activities (music, painting, gym,. . . ) 1Handicapped person or elderly, what advises to avoid stairs (yes/no) NoBulky objects to be stored in (number) 2

Fig. 7. Output: three two-stories solutions: [a2 1 3 4 7 111 206 20 5.7]; [a2 1 4 5 8 124 247 23.7 5.7]and [a2 1 4 4 8 118 226 20 5.7].


Note that the fourth to seventh entries could also be provided by a future extension of this KBS based onlocal town-planning codes, once the particular situation of the site is known or even inferred from GISinformation).

The KBS returns the best and second best options (in order), after processing the information. In thiscase:

(1) Three different house layout schemes of the type “rooms in a single line”, with service areas amongthem. This is the cleverest climatic reply, since it allows crossing airflows, and two-story buildingsare preferred as they may get upper flows (Fig. 7).

(2) Two other schemes of the type “rooms in a single line”, now with the service areas in a parallel line:not a very suitable type from a climatic point of view (and the system will warn of it), but none ofthe others could fit within a 7 m width building area (Fig. 8).

A first budget approach might be deduced from the size of the built area, the type chosen and the localcosts of materials and labor force. That is another future extension.

4.2. Another example

In case the input data were:

Building site area (m2) 2000Noise sources direction to dodge (no/yes: N-S or E-W) EBeautiful views to be considered (no/yes: direction N-S, E-W) NoBuilding rate 0.50Occupation rate 0.20Maximum height permitted (no. of stories) 3

Shape of the area able to be built on:Maximum length E-W (m) 25Maximum length N-S (m) 25

Basic climatic feature Temperate cold (with lowaverage hours of sun)

Number of people expected to live in the house 6Number of them working from home 2Number of special activities (music, painting, gym,. . . ) 1Handicapped person or elderly, what advises to avoid stairs (yes/no) YesBulky objects to be stored in (number) 0

Now the KBS gives priority to the only one-floor condition and returns:

(1) Two different house layout schemes of the type “rooms around a service core”, again a clever climaticreply, since these compact shaped houses would be kept warm easier in such a cold climate (Fig. 9).

(2) Two other layout schemes of the type “rooms around a court”: not a very suitable type from a climaticpoint of view (and, again, the system will warn of it), but maybe advisable for other reasons, like thepossibility of having an inner garden or no windows facing the noise source (Fig. 9).


Fig. 8. Output: two two-stories solutions: [e1 1 4 4 9 130 226 22 6.5] and [e1 1 4 5 8 118 206 18.5 6].

Fig. 9. Output: four one-story solutions: [u4 0 0 1 9 129 188 14.5 13]; [u4 0 0 2 11 149 234 18.5 13];[o3 0 0 3 6 128 219 18 15] and [o3 0 0 5 7 152 271 22 15].


5. Conclusions

This is a first development of a knowledge-based system for house layout selection. It shows howComputer Algebra Systems can be used as an aid in design processes. The users can consult the systemas many times as they wish, obtaining different solutions for each set of requirements.

We intend to include, as a future extension, construction elements real prices handling in order to obtainglobal budget computations.

At the present state, the system is constrained to single-family dwelling (houses), the most immediatelyapplicable to real life. Nevertheless, another interesting future extension would deal with multistorybuildings (apartment blocks).

Acknowledgements

This work was partially supported by the research project TIC-2000-1368-C03-03, Ministry of Scienceand Technology (MCyT), Spain. We would also like to thank the National Science Foundation (NSF),USA, for providing travel support to the first author.

We would like to thank Stanly Steinberg (University of New Mexico) for his most valuable comments.

References

[1] O. Luder, Great industrialized housing disaster, Building 247/42 (1984) 23.[2] C. Krause, Brain-storming context to improve industrialized construction, Architektur der DOR 39/2 (1990) 17–33.[3] G. Blachère, Technologies de la construction industrialisée, Éditions Eyrolles, Paris, 1975.[4] R. Bender, A crack in the Real-View Mirror: A View of Industrialized Building, Litton Educational Publishing, NY, 1973.[5] E. Levy, Industrialized building, Prog. Architect. 68/2 (1987) 92–95.[6] A. Klein, Vivienda mınima, 1906-1957, Gustavo Gili, Barcelona, 1980. Note: This a compilation of articles of the author

from 1926 to 1957.[7] R. Diamant, Industialised Building. 70 International Methods, Iliffe Books, London, 1968.[8] Association Construction et Composants (A.C.C.), Conventions de Cordinatione Dimensionnelle, Edit. Monitor, Paris,

1978.[9] A. Águila-Garcıa, Las tecnologıas de la industrialización en los edificios de vivienda, Colegio Oficial de Arquitectos,

Madrid, 1986.[10] G. Köhler, All quiet on the housing front? Detail 3 (2002) 172–178.[11] H. Nissen, Industrialized Building and Modular Design, Cement and Concrete Association, London, 1972.[12] C. Moore, G. Allen, D. Lyndon, Dimensions: The House, Shape and Design, Holt, Rinehalt & Winston, NY, Chicago &

San Francisco, 1974.[13] The American Institute of Architects, Regional Guidelines for Building Passive Energy Conserving Homes, AIA,

Washington, DC, 1980.[14] J.R. Goulding, J.O. Lewis, T.C. Steemers (Eds.), Energy Conscious Design, a Primer for Architects, B.T. Bastford Ltd.,

London, 1992.[15] E. Neufert, et al., Architect’s Data, Blackwell Science, Oxford, 2000.[16] A. Gonzalez-Uriel, E. Roanes-Lozano, Un Sistema Experto sobre Normas Urbanısticas, Bol. Soc. Mat. “Puig Adam” 61

(2002) 73–83.[17] URL: http://www.maplesoft.com.[18] M.B. Monagan, et al., Maple 9 Introductory Programming Guide, Maplesoft, Waterloo, 2003.[19] M.B. Monagan, et al., Maple 9 Advanced Programming Guide, Maplesoft, Waterloo, 2003.[20] R.M. Corless, Essential Maple, Springer-Verlag, New York, 1995.

http://www.maplesoft.com


[21] A. Heck, Introduction to Maple, Springer-Verlag, New York, 2003.[22] E. Roanes-Macıas, E. Roanes-Lozano, Cálculos Matemáticos por ordenador con Maple V.5, Rubiños-1890, Madrid, 1999.[23] J.A. Tajadura, J. López, AutoCad Avanzado 2002, McGraw-Hill Interamericana, Madrid, 2002.[24] J. Cuena, et al., Inteligencia Artificial: Sistemas Expertos, Alianza Editorial, Madrid, 1986.[25] G.F. Luger, W.A. Stubblefield, Artificial Intelligence. Structures and Strategies for Complex Problem Solving, Addison

Wesley Longman Inc., Harlow, 1998.

Documents

A knowledge-based system for house layout selection