The text presented here is not precisely as published by OUP, but modifications are minor. Illustrations are another matter. Where images used in the original book were not my copyright, I have in most cases been able to substitute links to coloured images on the web.

When this text was submitted as part of a PhD thesis in 1996, the Notes were greatly extended. Most readers may prefer to ignore them. They have been collected at the end of each chapter, with internal links leading to them and back to the text. They are a mixture of: simple page references; additional examples or quotations to justify generalisations; and some afterthoughts.

The nature and consequences of definition.

We shall now assume that the design process can be divided into three phases:

• Problem Definition and Clarification.
• Generation of Trial Solutions.
• Evaluation and Decision-Making.

We shall analyze each of these in turn and discuss some of the suggestions that have been made to improve their efficiency. However, the difficulty of separating the phases of design has long been recognized. The idea that in posing a question one necessarily implies a particular answer has also been current since Aristotle. [Note 1.] The separation into distinct phases is therefore artificial and in any discussion of one phase, the others must at least be mentioned.

As we have seen, as soon as a politician says 'we need a bridge across this river', he has, in specifying a bridge rather than a tunnel or a ferry, made an engineering decision for which he has no competence.

Insufficient attention has been paid to this phenomenon in structural engineering because the technique of 'definition by solution' is common practice. This is probably because most engineers are happy to avoid the difficulties of definition and prefer to be handed a well- defined problem which has been knocked into shape by someone else. Paradoxically such engineers are likely to be derisive about the contribution of those who have carried out this task because it must depend on possibly educated guesses of vague concepts and hunches which have yet to stand the test of investigation.

The way in which the client unit originally defines the problem will obviously aim subsequent efforts in a certain direction and will eliminate a vast range of possible solutions from consideration unless the design unit is sufficiently open-minded to see beyond the original brief, possibly in a field outside its expertise. The number of practicable solutions which are inadvertently overlooked will depend on the skill of the client unit in framing its brief.

Skilful definition of the technical problem is therefore a highly demanding task which can make an extremely significant contribution to the final design. The responsibility is all the greater because the initial vision has a compulsion which makes it seem more real than any which might arise later. For the reasons explained in the previous section it is far more likely that re-definition will proceed by modification of an existing proposal than by adoption of a radical alternative, even though the latter might be more appropriate.

The problem is compounded in conventional practice because client units rarely consider it necessary to acquaint others of even the conscious factors which affected their formulation of the problem. The consultant unit is thus unaware of many of the options directly open to it and cannot advise the client unit about them. Although the client unit must employ some discretion (otherwise it would not be fulfilling any function at all) traditional practice leaves much room for improvement.

Because of the cyclic nature of the design process, definition is a continuous process which extends through all levels. A study undertaken by the Tavistock Institute of Human Relations with the aim of improving on the RIBA Plan of Work found that

"… each time a design decision was taken it set in train a chain of consequences which could and did cause the initial decision to be changed … Since the full implications of any decision or action can seldom if ever be forecast with absolute accuracy, a communications system which assumes that they can will simply not work." [Note 2.]

Although this should be common knowledge many people in the industry behave as if it were not.

The definition of structural problems by architects is of special interest to structural engineers, and provides a good illustration of the principles set out above. As we have seen, the architect gauges the client's reactions to various proposals and points out to him the restrictive influence of such factors as finance, site conditions and material properties. He then conceives a geometry of form to satisfy the functional requirements as far as is practicable and in accordance with his aesthetic scheme. All this requires assumptions concerning the technical possibilities and hence some knowledge of structural engineering.

Even an engineer would have to depend upon fairly generalized concepts in making these assumptions. Unfortunately, many architects lack even such generalized concepts. If the structure is of a simple traditional nature, such as the average office building, no great problems arise. The structural form is stereotyped and the major feedback required from the design engineer is simply the sizes of columns and the depths of floor structures necessary to suit a given support system. However, where the structure is less conventional, the architect's arbitrary assumption of form may have drastic consequences, as we saw in the case of the Sydney Opera House roof.

Nevertheless it would be inefficient for the architect to acquaint the engineer with every minor consideration of economics, function, aesthetics and services which affected his decision. The structural engineer may therefore strike problems which could have been avoided if he had received more information. Opportunities may have been missed because the architect was not in possession of sufficient engineering knowledge, and the engineer's constructive suggestions for improved form may come to grief because of factors which the architect did not consider worth mentioning. The greatest opportunities for improvement, therefore, lie in maintaining good-will and open communication during this phase of design.

Many authors include with Definition the process of 'Clarification' or 'Detailed Analysis'. In doing this they are weakening the concept of Definition and incorporating more of the full design cycle. As the name implies, Clarification is the business of working out the ramifications of a given definition and checking for inconsistencies and internal conflicts. As cycling continues, the nature of the design space is determined, the basis for decisions becomes more precise, and the problem is defined in ever-increasing detail. The Definition thus becomes firmer, but so at the same time does the implied solution.

Common sense suggestions for improving definition.

Before we look at some of the formal, symbolic techniques which have been proposed for improving Definition, it is worth examining what can be done by simply taking stock of conventional methods.

The most effective means of improving efficiency seems to be to increase communication and understanding within the design team. Many architects now call in the engineer at an early stage for mutual involvement in planning while others have formed inter-disciplinary teams. This is a welcome development.

The need for improved communication exists not only across the boundaries of disciplines but also between the different levels of hierarchy within organizations. The new graduate is more likely to be briefed by a senior engineer than an architect. Senior engineers could do much by taking their subordinates more into their confidence and being more receptive and tolerant towards their suggestions.

It is at the stage of feedback that the junior engineer has his greatest opportunity to make a real contribution to the design. Intelligent communication of the problems encountered and suggestions for avoiding these may assist in achieving beneficial re-definition.

How, then, should he go about this? To achieve a good design he should ideally determine the factors which motivated the formulation of the problem and ask himself: 'If those who posed the problem knew as much as I do about the possible answers, would they still have posed it in that way? What do they really want?'

As we have already seen, behind every expression of perceived 'need' lies a more broadly definable motive. When a senior engineer says to his section head 'design me a superstructure for the machine hall of this power station' his underlying motive is to provide support for an overhead crane and to protect the generators and turbines from the weather. An alternative, though much less common solution, is to provide an independent travelling gantry crane and to weatherproof the machine casings individually.

It is a fine point to what extent a designer, particularly a young engineer should investigate the formulation of a problem by his 'client unit'. Obviously when his 'client' is a senior engineer the chance that the designer will be able to make a positive contribution to the redefinition of the 'problem' (and the chance that it will be appreciated) is less than if his client is an architect with a commitment to structural rationalism. However as he tackles his appointed task he will develop an intimate knowledge of a part of the 'design space' that others cannot possibly have had in advance.

There must, of course, be a limit to investigation and redefinition and the depth of inquiry in any particular circumstance must depend on the individual's assessment of its value relative to the time available and the willingness of superiors to tolerate the process!

Some of the questions a designer might ask are:

• Has the client unit, in communicating its vision of the problem, implied a structural solution which is inappropriate to its aspirations or is unnecessarily costly?
• Are any of its aspirations mutually exclusive for technical reasons?
• Will any aspiration of minor importance cost a disproportionate amount of money?
• Will the implied range of solutions have disadvantages or adverse effects unforeseen by the client unit?

If affirmative answers to any of these questions emerge during the process, the information obtained should be fed back to the client unit.

A major ethical problem arises when the client resists the designer's attempts to widen the enquiry on the grounds that he 'knows what he wants' and does not wish to pay a higher fee for investigations he considers to be unnecessary. An example of this is the client who asks for stress analyses of structures or components for which he specifies highly simplified loading or boundary conditions. If the consultant realizes these are totally unrepresentative of the actual situation he should withdraw his services rather than accept a fee for a meaningless calculation.

The engineer should use particular caution when a layman couches a brief in highly specific engineering terms, particularly when this involves the modification of a part of an existing structure. In this case the client's ignorance of the significant engineering factors may be positively dangerous.

Formal symbolic methods for improving definition and clarification.

It is indicative of the problems of separating phases that Jones (1981) places many of these formal methods for Clarification in his middle phase of synthesis or "transformation".

During the process of Definition a list of the "elements" of the problem is drawn up. One way of doing this is to prepare a "Morphological Chart". The first step is to "define the functions that any acceptable design must be able to perform" and the second to "list on a chart a wide range of sub-solutions, i.e. alternative means of performing each function". These may be the rooms required in a building, the functions required of a given room, or the target tasks of a crane or mine head-frame. Note that the creation of such a list is in fact the invention of a list of particular sub-solutions to the sub- problems.

The next stage in formal methodology is to note the interrelationships between these elements. If they are tasks, it may happen that a possible sub-solution will satisfy several requirements simultaneously. On the other hand, it may satisfy one and interfere with another.

Fig. 14.1. (a) In this interaction matrix the component spaces of a house are listed on both axes. On either side of the main diagonal are noted the levels of desirability of aural and visual communication. (b) The resulting information has been transformed into an interaction net which will assist in planning the layout of spaces. (After Broadbent 1971.)

Formal methods of indicating these relationships on paper include the 'Interaction Matrix' (Fig. 14.1) and the 'Interaction Net'. Methods based on matrix formulation include AIDA (Analysis of Interconnected Decision Areas) and Alexander's 'Method of Determining Components'. [Note 3.] In AIDA certain reasonably coherent elements are defined. In the design of a house, these might be the roof, walls and floors. Possibilities regarding the nature of each of these are listed and an interaction diagram is drawn to show where any one of these characteristics (which must be mutually exclusive) will affect decisions regarding other elements. The links thus represent incompatibility between options, which is the reverse of the convention adopted in most schemes. A typical net diagram is shown in Fig. 14.2.

Fig. 14.2. An AIDA net in which characteristics of architectural components have been listed and incompatibility indicated by connecting lines. (After Broadbent 1971.)

As we have seen, these methods superseded earlier ones which such defined and clarified problems by breaking them down into a number of sub-problems which were basically independent of one another. The symbolic representation of this is a tree diagram. Where interaction was seen to occur between elements, it was assumed to be of a very simple kind; the requirements either contradicted one another or they did not. Later ideas took account of the fact that sub-problems and their solutions can rarely be separated in this way and are better represented by a semi-lattice (Fig. 14.3).

Fig. 14.3. The semi-lattice, an extension of the tree-diagram concept which recognises that problems can rarely be broken down into entirely independent sub-problems.

A major disadvantage of such representations is that it is difficult to indicate the degree of importance of relationships. Furthermore, the number of possible solutions rises enormously as the number of variables is increased and it is necessary to use a computer to search for and select what are apparently the best solutions.

The only way in which the present generation of computer programs could handle qualitative rather than quantitative relationships would be by allocating relative weightings to the various factors (e.g. 'ease of access' is worth three times 'aesthetic merit', while brick has twice the aesthetic merit of concrete). Such sweeping generalizations would be treated within the program as having universal application and truth, whereas of course they would be considered by a human designer to vary with each individual situation.

As a result current computer applications tend to concentrate on one particular aspect, such as optimizing circulation patterns within a building and the result must then be modified by the designer to take account of the factors which have been overlooked. The designer then still does not know how many possible satisfactory solutions close to the global optimum have been overlooked.

For these reasons, formal design methods have aroused little interest amongst structural designers, and it is too early to say whether they have great potential in this field. Judging from experience in other areas, it is likely that their main beneficial effect will be in changing the attitude with which designers approach their work.

Notes.

Note 1. Choisy, the great French professor of architecture taught around the end of the 19th Century "la question posée, la solution est indiquée". (Cited in Banham 1960, p.26.) [Return.]

Note 2. Cited in Broadbent (1973), pp. 268 and 269. [Return.]

Note 3. Broadbent (1973) pp. 280-3 and 274-80. [Return.]

[Top.]

The art in structural design navigation:
[Chapter 13.] [Chapter 15.] [Contents] [References]

Site navigation:
[Aesthetics of built form.] [Papers.] [Work of Jörg Schlaich.]
[John Monash's early engineering.]