![INTERNATIONAL CONFERENCE ON ENGINEERING DESIGN
ICED 01 GLASGOW.AUGUST21-23, 2001
IMPROVING ACCESS TO DESIGN SOLUTION SPACES USING
VISUALIZATION AND DATA REDUCTION TECHNIQUES
P M Langdon and AChakrabarti
Keywords: Visualisation, clustering, synthesis, design information management, information
analysis, classification and retrieval
1 Introduction
Designers only explore a few solutions in depth at the conceptual stage [1]. Despite this,
evidence suggests that a thorough exploration of a solution space is more likely to lead to
designs of higher quality [2]. FuncSION [3] is a computational tool that synthesises a wide
range of solutions to a class of mechanical design problems involving transmission and
transformation of mechanical forces and motions that are specified as inputs and outputs. It
uses a set of primary functional elements along with combination rules to create anexhaustive
set of solutions in terms of their topological and spatial configurations. Previous research has
demonstrated FuncSION's potential for generating novel solution ideas that designers had not
thought of. However, it was found that the large number of ideas generated could not be
meaningfully explored, while their representation proved too abstract to visualise. Effective
support for conceptual design should help designers obtain a thorough overview of the
solution space, as well as a detailed understanding of its individual solutions. However, the
greater the variety and number of solutions to be explored, the less likely it is that a detailed
understanding of the potential of all individual solutions will be achieved. The DESYN
software described in this paper encapsulates FuncSION with a Graphic User Interfacce
(GUI) [4]. The overall aim of this work is to test the extent to which DESYN is capable of
assisting designers in their exploration of large solution spaces such that an overview of the
entire space is obtained while at the same time facilitating understanding of the individual
solutions contained in it.
2 Background
A paper presented at ICED99 [4] described a scheme adopted in order to solve the above
problems. The problem of exploring large combinatorial spaces is an unresolved issue in
computer aided synthesis research, which is further complicated by the difficulty in displaying
large volume of information [5].
The approach adopted was a novel method of clusteringthe solution configurationsto reduce
the space of solutions that the designer needs to consider in order to get an overview of the
entire space of solutions [6]. This was done by presenting the designers with representative
solutions that were by-products of the clustering process. In this way, large numbers of
solutions can be summarised by a small number of cluster exemplars of prototypes.
ICED 01 - C586/223 3](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-23-320.jpg)
![A number of previous approaches have adopted rule-based heuristics for the generation of
solution compositions in an unconstrained space (See, for example, Campbell et al, 1999)[7]
[8][9]. However, the approach adopted here assumes exhaustive generation of solutions in a
partially constrained space with the use of clustering as a data reduction and representation.
Because of this, all solutions are, in principle, available to the designer through interaction
with the systems GUI.
In the authors' earlier paper, some initial validation of the cluster algorithm was reported
focussing on whether the system's clusterings of solutions were intuitive and whether the
clusterings suggested by the system correspond to designers' own partition of the solution
space. Initial experiments examined a number of DESYN clusterings of two sets of 20
solutions resulting from a synthis using 0-5 elements per solution from a choice of < 4
elements. Figure 1 is an example diagram of the outputs of the same clustering algorithm for a
smaller 6 solution set synthesised using the elements wedge, cam and lever mechanisms.
Clustering was based on a Euclidian distance metric operating on a count of features in each
solution from a feature set of all possible feature pairs.
Figure 1.Cluster table for 2 to 5 clusteringsof a 6 solutions for a problem using Wedge, Cam and Lever
elements.
The arrows indicate the solutions that leave the 3 cluster solution to form new clusters. On the
left of the diagram the six possible solutions output by the synthesiser are enumerated. The
vertical columns show increasing number of clusters in the solution set. The cells show the
cluster membership for each solution. Each cluster has a representative solution that is
denoted by circling. Finally,the box in the corner of cells shows the average distance of the
representative to all the other members of the cluster. The right hand diagram shows a Venn
type representation of the 3,4 and 5 cluster solutions. Solution sets for the larger 5 element
problems were presented to 5 subjects as printed words with illustrative diagrams.
Subjects were required to form groupings of the 20 solutions that corresponded to their
judgements of those that seemed to them to be similar on the basis of their engineering
experience. The subject's groupings were scored for the percentage similarity of groupings
with the DESYN clusterings of the same set. This was to test whether designers found the
system's clustering of solutions intuitive, and whether the clustering suggested by the
designers correspond to their own partition of the solution space. This clustering was based on
an element-pair feature count (focusing on the type of interfaces possible as design elements
in combination),and provided an average commonality, between designer's clustering and the
4 ICED 01 - C586/223](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-24-320.jpg)
![method's, of 74% in contrast to 55% commonality yielded by comparison with a random
clustering.
However, this evaluation was carried out using only the topology, rather than the 3-D
configurations of the solutions. Evaluation of the approach is further extended in this paper
(see section 3.2) using: (1) 3-D configurations displayed by the system as displays of force
and motion solution chains; (2) Graphic representations of the clustered solution space. The
goal of this evaluation is to find whether or not this clustering-based technique improves a
designer's overview of the entire solution space, and whether the software facilities assist
their understanding of the solution space.
3 Current Developments
Current developments include further development of the GUI and visualisation support, and
further evaluation of DESYN, which are described below.
3.1 User interface development
The visualisation utilises a 2D 'star' representation of clusters (Figure 2), and a pseudo-3D
display of a component chain schematic (Figure 3) that we have developed to improve access
to the functional synthesis software. The former is used to aid visualisation of the spatial
relationships between the component interfaces, while the latter is intended to aid the
understanding of the grouping of functional solutions by similarity [10, 11].
Figure 2. The DESYN Graphic User Interface Clustering Tool
In conjunction with this, a 3D visualisation for displaying an overview of the solution space
and the spatial layouts of selected solutions resulting from the synthesis is under development.
ICED 01 - C586/223 5](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-25-320.jpg)
![the screen. No indication of the total number of solutions was given and the solutions were
not presented in any ordering linked to their functional structure. Hence, the designers were
allowed to select as few or as many solutions as they liked. The task further required that the
designers selected individual solutions form the list (left, Figure 4) that they felt met the
criteria. This was made on a point-and-click basis and the resulting set stored in a further text
field (right, Figure 4).
Figure 4. The list representation experimentalgroup interface.
In the structured condition the designers performed the same task but interacted with a 2D
cluster representation of the solutions set. Dwelling the cursor on a central representative
"sun" member of a cluster lead to a list of the cluster solution members appearing in the upper
right text field. Dwelling on the numbered "planet" members highlighted the location of the
solution in the list.
Figure 5. The list representation experimentalgroup interface.
Again, the task further required that the designers select individual solutions form the list (top
right, Figure 5) that they felt met the criteria. This was made on a point-and-click basis by
clicking on the member "planets" and the resulting set stored in a further text field (bottom
right, Figure 5). In both conditions the designers were able to delete members of their
selected set at will.
3.3 Data analysis technique
Results in our previous study [4] suggested that the clustering solution for small problems had
some psychological validity as a grouping of the solution space correlated highly with
designers groupings. Therefore, in the present case, if a solution chosen by a designer
belonged to a cluster generated by the computer, then it was assumed that the designer had an
overview of at least that cluster. If there was a better match between the computed clusters
ICED 01 - C586/223 7](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-27-320.jpg)
![tested in these experiments. However, it was observed that the time to complete the task was
significantly shorter (about 50%) in the clustered condition. Although this was not measured
accurately, it may suggest that the list condition designers required more time to gain an
overview of the space than the designers for whom the solution space was spatially laid out.
The overwhelmingly popular solution represented three sub-mechanisms that give key two
element combinations that re-occur. These included mechanisms that allowed orthogonal
changes in force direction, such as lever to lever; lever to cam; or lever to screw links. The
reports collected from these designers suggested that the criteria applied was simplicity of
solutions, ignoring repetitions of sub-mechanisms and spatially translating force to force
elements such as tie-rods. The effect of prompting the designers for more solutions in these
cases led to the choice of variants on these mechanisms with tie-rods added to circumvent
possible spatial constraints.
The limitations of the evaluation lie in the small number of designers sampled and the use of a
single problem. In addition the previous evaluation [2] had established a commonalityof 75%
between the software-clustered groups and the designers intuitions as indicated by a paper-
based solution-sorting task. In addition, the clustering method was based on a feature based
on a count of pairs of solutionelements. Other features of the solution chain could be used for
clustering to further test the effectiveness of the clustering approach. It was also clear from
the designers performance that a more realistic task would involve a realistic design task
carried out over a longer period. In addition to this, the pseudo 3D visualisation software was
not integrated into the DESYN interface such that visualisation of individual solution chains
was possible during the trial. It was evident that such a visualisation tool would have assisted
the designers understanding of the individual solutions. Work is currently underway with a
larger group of designers, sampling a wider range of problems over a range of sizes of
mechanisms and difficulties of task.
6 Conclusions and further work
A software system for assisting the visualisation of alternative solutions to mechanical design
problems involving transmission and transformation of mechanical forces and motions was
successfully implemented. This utilised a clustering algorithm for the purpose of data
reduction and visualisation of a large solution space. An empirical evaluation of this system
examined whether the designers' sampling of the known solution space was effectively
assisted by an element-pair clustering represented in a 2D display of clusters, in conjunction
with a pseudo-3D visualisationtechnique.
There was no evidence for any improvement in the number of clusters sampled when the
interface provided a list of solutions as opposed to a 2D cluster membership diagram, though
a substantial improvement in time to complete the task was noted. Two strategies were
observed in the designers tested. The predominant strategy involved listing 'elemental' sub-
mechanisms that were capable of orthogonal changes in force direction. These designers
ignored repetitions of sequences or complex combinations as well as elements providing
linear or parallel force transformations. These second strategy group appeared to list solution
sets that were clearly distinctive sequences, because of simplicity or unique element
combinations.
The small number of designers sampled does not enable generalisation to designers at large as
the results may reflect individual strategies. It is also possible that designers' choices were
ICED 01 - C586/223 9](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-29-320.jpg)
![affected by the lack of a genuine engineering task in the short trials. Firmer conclusions will
be facilitated by the collection of further data using an enhanced assistive interface and more
realistic and rigorous design tasks.
7 References
[1] Chakrabarti, A. and Wolf, B., Reasoning with Shapes: Some Observations from a Case
Study", Proc. 1995 ASME Design Engineering Technical Conferences: (9th Intl. Conf.
on DTM), DE-Vol. 83, Vol.2, pp-315-322, 1995.
[2] Fricke, G., Experimental investigation of individual processes in engineering design,
Research in Design Thinking, (N.Cross, K. Doorst and N. Roozenburg eds.) Delft
University Press, pp 105-109, 1992.
[3] Chakrabarti A., and Bligh, T.P. "An Approach to Functional Synthesis of Design
Concepts: Theory, Application, and Emerging Research Issues", AI in Engineering
Design, Analysis and Manufacturing, Vol. 10, No. 4, pp-313-331, 1996.
[4] Langdon, P., and Chakrabarti A. "Browsing a large solution space in breadth and
depth", A. 12th International Conference on Engineering Design (ICED 99), Munich,
Germany, v3 p. 1865-1868, 1999.
[5] Johnson,B., & Shneiderman, B. "Tree-maps: A space-filling approach to the
visualisation of hierarchical information structures". Proc. IEEE Visualization'91.
IEEE, Piscataway, NJ, 284-291, 1991.
[6] Kaufman, L. & Rousseeuw, P.J., Finding Groups in Data. Wiley Series in Probability
and Mathematical Statistics. John Wiley and Sons, Inc., 1990.
[7] Campbell, M.I. ,Cagan, J., Kotovsky, K., A-Design: An Agent-Based approach to
conceptual design in a dynamic environment. Research in Engineering Design, Vol. 11,
pp 172-192, 1999.
[8] Finger, S., and Rinderle. J.R., A transformational approach to mechanical design usinga
bond-graph grammer. In Design Theory and Methodology, DTM 89, vDE Vol. 17. pp
107-115, 1989.
[9] Bracewell, R.H., Sharpe, J.E.E. Functional descriptions used in computer support for
qualitative scheme generation - Schemebuilder. AI EDAM Journal - Special Issue:
Representing Functionality in Design 10(4): p. 333-346, (1996).
[10] Robertson G.G., Card S.K., Mackinlay, J.D., Information visualization using 3D
interactive animation. Communications of the ACM, Apr 1993, Vol.36, No.4, pp.57-71,
1993.
[11] Mackinlay,J.D., Rao.R, Card, S.K., An Organic user interface for searching citation
links. Human Factors in Computing Systems (CHI) Conference Proceedings, Vol.1,
pp.67-73, ACM, New York, NY, USA. 1995.
Corresponding author's name: Dr. Patrick Langdon
Engineering Design Centre, Cambridge University Engineering Department, Trumpington
Street, Cambridge CB2 1PZ, UK, Tel: +44 1223 766961 Fax: 444 1223 332662, E-mail:
pml24@eng.cam.ac.uk, URL:http://www-edc.eng.cam.ac.uk/people/pml24.html
© Cambridge Engineering Design Centre2001
10 ICED 01 - C586/223](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-30-320.jpg)
![INTERNATIONAL CONFERENCE ON ENGINEERING DESIGN
ICED 01 GLASGOW, AUGUST 21-23, 2001
SUPPORTING NON-STRUCTURED GRAPHICALINFORMATIONIN
INTEGRATED DESIGNTEAM
E Blanco and M Gardoni
Keywords: design information, design understanding, hypermedia and multimedia,
knowledge management
The aim of this paper is to focus on the use of draft and sketches in the group understanding
within Integrated-Team. We suggest specifications of a communication tool that supports and
structures numerical drafting and treatment for capitalisation of those drafts connected to
messages.
1 Evolution of Engineering Information Requirements
Traditionally, engineering activities are performed in a sequential order. Over the last ten
years, companies have tended to apply the Concurrent Engineering approach (CE) in order to
drastically reduce the time-to-market of their products [1]. This approach is opposed to the
sequential engineering approach because they have respectively two fundamental ways of
working. In sequential engineering approach, the work starts at the reception of the results of
the early stage when CE is based on information exchanges. Those exchanges are mainly
performed verbally face-to-face or on the phone. Unfortunately, this information is therefore
poorly controlled. This term "control" can be characterised in terms of four criteria [2] :
- Information Structuring: it relies on a formal conceptual scheme which organise
information at appropriate places. The evaluation criteria are the easyness to organise
information in an intuitive and logical way and to be able to manage information at
separate locations.
Information Sharing: ability of "pushing" information.
- Information Access: ability to "pull"information.
- Information Capitalisation: ability to store and process information for later re-use, we
assume that it must comply with the cycle of 'company knowledge capitalisation', i.e.
locate, memorise, use information and update information.
Taking into account these new requirements, we characterise the type of information
exchanged in Integrated Team. We will focus here on information Structuration model. In
order to meet the rigour needs of companies without going down on a too fine level of
granularity, we choose an instructional design of the information significance. We then
consider that the construction of a sentence corresponds to combine instructions formulated in
term of variables, which provide a sense to the statement. Exchanged information is then an
abstracted entity, a theoretical object which consists of [3]:
ICED 01 - C586/418 11](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-31-320.jpg)
![- Linguistic components which build the significance of information starting from
instructions.
- Rhetoric components which bring a sense to information by addition of contextual
information.
Figure 1. Instructional design of the information significance
Thanks to this instructional design of the information significance, we characterise different
types of information [2] :
Structured-Information (SI), linguistic and rhetoric components of the 57 are generally
imposed.
- Semi-Structured-Information (SSI), linguistic components of the SSI are little formalised
and rhetoric components could be parsimonious.
Non-Structured-Information (NSI), the NSI are very little formalised and the rhetoric
components can be very light if they ensure a sufficient degree of relevance for the
comprehension of information by the receiver.
The MICA approach [4] (a specific interactive messaging system) has been developed and
experimented in order to harness of NSI and to capitalise on relevant information exchanges.
Also a Groupware tool called MICA, was created through an Intranet. It has been
implemented and put into operation in an engineering team of twenty people in May 1998.
Some return on experience has already shown improvement of the efficiency of the Integrated
Team.
Nevertheless, this kind of capitalisation from linguistic data is limited because of the
graphical NSI lack. Sketching takes a large place in the group understanding of technical
problems. Goel [5] also argues that freehand sketches facilitate creative and explorative work
at the early stage of design, because they are ambiguous semantically and syntactically dense.
Sketches could be considered as Graphical Non Structured Information (GNSI). They are at
the same time models of the product and communication vectors. MICA only takes into
account linguistic data but does not take care of GNSI management. This is today one of the
actual limit of the MICA approach and more generally of Knowledge Management systems
[6]. Generally Groupware solutions for interactive sketching have to be improved.
2 GNSI in collaborative design
In order to analyse the role of GNSI in collaborative design we have carried out a design
experiment [7]. This experiment was based upon the distributed design model [8]. Three roles
12 ICED 01 - C586/418](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-32-320.jpg)
![were represented, the functional, structural and machining roles. A camera has recorded the
whole progress of the design. The work began with the requirement list supplied by the client
a few days before the experiment. It took six hours to implement the detailed drafts of each
part of the product in order to manufacture it.
2.1 Sketches as IntermediaryObjects of the designprocess
In order to observe and analyse the design processes, we suggest to start with the Intermediary
Objects (IO) which, circulating at any given moment between the actors of the process, can be
seen as resulting from their design work but also as supporting and highlighting it [9], [10].
In modelling the future product they also act as communication vectors between the product
designers. These two aspects are so indissociable in the reality of the process that we cannot
isolate one from the other without deforming their nature. Because of this hybrid nature,
intermediary objects are "analysers", making it possible to describe the actual design process.
All along the design task, the object can not exist without the actor and the actor without the
object. What we observe, see and feel in the design process, are objects being constructed,
talked about, manipulated, interpreted, transformed, etc. Considered separately, the objects
and actors are merely capacities for action. They are inert, static, mere?? "possibilities". It is
the action, or rather the inter-action in which they are engaged that endows them with force,
meaning and effective reality.
The intermediary objects are also mediating objects: first of all, because of the mode of
representation they use. This leads to a particular way of objectivising the idea or the intention
by inscribing it in a specific organised matter. Thus, for example, a drawing is not simply a
faithful representation of the mental idea or specifications : it is a translation, i.e. a realisation
and a transformation which has been carried out according to its specific constraints
(including those of the material used, which may vary depending on whether it is on paper or
on screen) and its own rules and conventions. Lastly, the way in which the Intermediary
Objects play the role of mediators in the design process is a result of their being
representations.
IO contents are mainly of cognitive nature : it results from the use of acquired knowledge by
the actors in their various fields. They also signal the gradual production of new knowledge
about the product as it is being designed. However, these contents are only of a cognitive
nature in a situation of action, oriented by a given project and made up by the choices of the
actors, the multiple negotiations that they are involved in, and the compromises and decisions
that result from these. Therefore the cognitive contents cannot be isolated from the context of
action
2.2 Classification of sketches in the collaborative process
The role of the objects like sketches in collaborative design process are quite complex.
Ferguson [11] identifies three kinds of sketches : thinking, talking, prescriptive. The situation
of the experiment that we have carried out emphases on the second category of sketches. The
collaborative activity involved many talking sketches. On the contrary, studies based on
single designer activity mainly point out the thinking role of sketches [12].
In the experiment, we had characterised sketches [6] through an axis from open object that are
able to support negotiation, to closed objects that mainly support prescriptions. We observed
ICED 01 - C586/418 13](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-33-320.jpg)
![that the same object can support different status even if some objects get materials and
cognitive characteristics that enhance opened or closed use. The situation of action as well as
the material and cognitive properties of the IO influence the status.
In the protocol, for example, we observed that the same sketch can be realised by an actor for
his own use and after a few minutes be proposed to the others for evaluation. It moves from a
private status (it is a thinking sketch), to a public one in the centre of the table. The existence
of the private area is very important for the designer. This area is the thinking area that allows
to explore ideas without judgement from the other. In a second step, the actor presents the
meaning of his sketch and the other actors can assess the solution in their own point of view.
This object then acts as a conjecture of solution. We can say that it is an externalisation of the
designer's solution mental image. But this talking phase is also a thinking phase for the group
that can make the sketch evolve. Every sketches are not built in a private area. Some are
drawn directly in the public area of the table.
A Sketch can also move from a talking to a prescriptive status. A sketch which has been the
instrument of solution negotiation and elaboration, can get a prescriptive status after the group
has built an agreement over it. For example, the sketch figure 2 got this prescriptive status by
the group agreement. This decision is confirmed by a mark on the object: One of the actor
underlined the word piston. This mark (detail Al) acts as the validation of the agreement. The
definition and the negotiation of this part is then closed.
Goel [5] suggests another way to characterise the evolution of sketches in he design process:
the transformations typology. He identifies two types of operations occurring between
successive sketches in the early stages of design: lateral and vertical transformations. A lateral
transformation movement goes from one idea to a slightly different idea. Vertical express a
movement from one idea to a detail or an extract of it.
Then we can say that the role of the sketches depends on their status and the evolution
between to successive sketches. The next table sum up the different categories of sketches we
have identified.
Thinking
Talking / open
Prescription / closed
Lateral transformation
Private new conjecture
Public new conjecture
New prescription
Vertical transformation
Private in depth conjecture
Public in depth conjecture
In depth prescription
The status of the sketch is important in the sense that they partially characterise the situation
of action where the actors are involved. Referring to the model presented figure 1, we can
notice that the perception of the situation allows the receiver to build the sense of the
information within the situation S. It act as a part of rhetoric components.
14 ICED 01 - C586/418
Table 1. Typology of sketches
Private area
Public area](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-34-320.jpg)
![2.3 Sketches as a process of building conventional support
For the explanation of this point, we must focus on the building process of those objects. The
Object presented in figure 3 is in fact an addition of different sketch steps. The building of
this sketch represents 20 minutes of group interaction during the design process. We consider
that it is in this object that emerges the solution principle developed during the next phases of
the experiment[7]. The first sketch (detail A) was drawn to put a end to a misunderstanding
between two actors. They did not have the same interpretation of the requirements. The three
sketches (A, B, C) allowed them to build a shared understanding of the problem. Then, from
this representation a third actor had doubt about the connection between the device and its
environment. This actor assessed the rubber tubing connection (detail D) as a non reliable
solution for this system. From this negative evaluation of the solution elements represented on
this sketch emerged a few minutes later a rigid and quick connection solution that is
represented (detail E) on the sketch. This move is a lateral transformation that highlight the
emergence process.
Then this solution offered new opportunities for the inter-connections system of blocks.
Finally a general overview of the system was drawn to visualise the ability of blocks
interconnection (detail F). This last view is a vertical transformation showing the same
solution from another point of view.
Figure 2. Status of object moves during the
process.
Figure 3. Different level of sketches in the same
draft...
This sketch shows that the knowledge of construction process is necessary to be able to
understand the sketch. In this sketch, aborted solutions and the solution chosen are
represented on the same level. Nothing but the knowing of the process allowed the actors to
differentiate the good from the aborted solution.
This point allows us to point out that this type of GNSI is not able to support a prescription
task. It failed as a memory of the process, In our protocol, we highlight that the actors
themselves had difficulties to re-use their own sketches out of the action process itself.[7]
The actors created, during the process, some symbolic devices which were connected to
semantic in a local convention. This allowed them to let some part of the device in a low
definition level: fuzzy and partial. The objects were used as "cognitive artefacts". The
participants did not describe precisely what they were talking about, they showed the different
elements they wanted to highlight with a finger or a pen and they used diectic words to point
them. For example, an actor said: "that will be difficult to manufacture" (showing the
elements concerned with his pen)
ICED 01 - C586/418 15](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-35-320.jpg)
![We want to highlight the fact that the sketches act as pragmatic conventional supports [13].
Those conventional supports are negotiated in the interactions between actors, even if they
can also use a higher level of convention as cultural ones, shared by the actors (the rules of
industrial drawing in our case).
The sense of the sketch is built in the action process and the conventions which allow this
interpretation are quite local. Rhetoric components of sketches are specific and part of the
linguistic components are built within the action itself.
This characteristic of GNSI is important in order to imagine design tool able to support GNSI
in an asynchronous communication process; furthermore, in order to involve sketches in a
knowledge capitalisation process as MICA does for textual NSI. Indeed local conventions
which allow interpretations are built in the interaction process. They won't be available to
actors that were not involved in the process itself.
3 Towards a communication tool MICA GRAPH
The first quality of a sketching device for designer seems to allow fast freehand sketch.
Previous studies have highlighted that the cognitive tasks of manipulating the sketching tool
have to be minimised [14]. Even if we can expect that the actors will learn the tool. That is
why the tool is based on a simple blackboard technic and uses a digital table with an
electronic pen. Indeed, when the designer needs formal drawing he will mainly use CAD or
Structured information tools. Sketches are just GNSI and support fuzzy and partial definitions
of elements. The functions that we offer are not supposed to help the drawing process but to
partially structure GNSI in order to capitalise and treat them in the knowledge process. We are
at the moment unable to manage a knowledge treatment directly on sketches. The principle
we develop here is to characterise GNSI, track the steps of the building process and to
associate textual and symbolic information that the data-mining techniques used by MICA are
able to treat. At the moment, a designer creates a new sketch, he creates a properties file
containing the legend of the file, type of the GNSI created and all the textual annotations
contained. All this textual information is able to be treated by the MICA data-mining
process [4].
3.1 Tracking the different steps of the object
We have highlighted the difficulty to track the different steps of a sketch out of the drawing
process. It is also impossible to identify in a sketch the valid elements from the aborted
solutions. It is important to follow the sketches modifications and to track the different sketch
levels and steps present in the same final draft. We also want to know who had drawn on it.
We propose to develop a structure of layers in order track the process. Each actor who wants
to draw on an existing draft has to open a new layer. He can chose two options:
"transparency" that allows him to sketch on the previous element or "blank" that opens a new
layer disconnected from the previous sketch. In order to track information about the sketch
contents and intentions, we characterise the type of opened layers. Six types of layers are
available corresponding to the typology proposed in table 1. Public layers and private layers
have different properties. When an actor wants to send a sketch, the public layers are
automatically sent. Private layers sending is optional. If the actor accepts to send them the
category changes moving from private to the public area.
16 ICED01-C586/418](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-36-320.jpg)
![In order to highlight decisions concerning a specific area of a sketch, actors can modify the
colour of this area. By drawing a boundary line around this area, they notify that it is fixed.
This materialises the compromise within the object.
3.2 The necessity to associate annotations on GNSI
In order to allow the actors to interpret the sketches in an asynchronous communication mode,
and to capitalise textual information about the content of the sketch, we propose different
devices to the users. Firstly an actor as to define a title to the sketch using few keywords. We
notice here that some contextual components (product, process, state, part etc. ) are already
identified in the MICA form.[2] [4]
It is important to give the actors the ability to express comments in the sketch itself. In a
synchronous interaction around sketches actors explain some elements of the solution they
focus on. Drawing is generally accompanied by a speech concerning the sketch. In an
asynchronous process, the speech does not exist. It is supported by writing in the message or
by a local annotation on the sketch.
In the same way, actors need to be able to focus the attention of the receiver on a specific area
and to associate elements that allow the receiver to interpret the sketch in the right way. They
have sometimes to explain some symbolic elements they have used to fuzzy represent an
element. Annotations can allow them to do it. The annotations are textual information that are
connected to the graphic properties.
A further step is to allow an actor to define different types of annotations using colour and
type of markers. Each time he defines a new type, he has to explain the legend of this
symbolic feature. Laureillard had pointed out the importance of what he called co-operative
feature in the co-operation between specialists [15][16]. Those symbols are local conventional
supports which refer to shared knowledge between the different actors. We don't have deeply
investigated the different types of annotation. Our proposition is to offer a toolbox to the
actors. They have to define themselves the relevant type co-operation feature.
4 Conclusions
The concepts developed in the MICA-GRAPH tool have to be improved in design practices.
The aim is not to suppress synchronous interactions but to complete the offer of asynchronous
communications tools for design activities. Engineering design needs visual representations:
structured (like CAD models) and non structured (like sketches). GNSI allows to manage
quick conjecture-evaluation process. We notice that the move from direct interaction to
textual explanation implies the risk of the deterioration of the contents. But the advantages of
textual information related to GNSI aretheir ability to be treated in a knowledge process.
References
[1] B. Prasad, Concurrent Engineering Fundamentals - Integrated product and process
organisation, Vol. 1, Prentice Hall, Englewood Cliffs, NJ, 1996
[2] M. Gardoni, M. Spadoni, F. Vernadat, "Information and Knowledge Support in
Concurrent Engineering Environments", 3rd
International Conference on Engineering
Design and Automation, EDA'99, Vancouver, B.C., Canada, August 1-4, 1999
ICED 01 - C586/418 17](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-37-320.jpg)
![[3] Moeschler, J. Modelisation du dialogue (representation de 1'inference argumentative),
Editions Hermes, Paris, 1989
[4] M. Gardoni, Maitrise de 1'information non structuree et capitalisation du savoir et du
savoir-faire en Ingenierie Integree - Cas d'etude Aerospatiale Matra, PhD thesis of Metz
University 1999
[5] Goel V, Sketches of thought MIT press Cambridge, MA 1995
[6] Gardoni M., Blanco E Taxonomy of information and capitalisation in a Concurrent
Engineering context 7th ispe international conference on concurrent engineering
(CE'2000), Lyon, France, July 2000
[7] Blanco E., 1'emergence du produit dans la conception distribute PHD thesis INP
Grenoble 1998
[8] Garro O., Salau I., Martin P., Distributed design theory and methodology, Concurrent
engineering: research and applications, vol 3/1/1995.
[9] Vinck D., Jeantet A., 1995 Mediating and commissioning objects in the sociotechnical
process of product design: a conceptual approach, pp111-129 in D. Mac Lean, P.
Saviotti, D. Vinck (eds), Management and new technology : Design, Networks and
Strategy. COST Social science serie. Bruxelles. Commission of european
[10] Blanco E., Garro O., Brissaud D., Jeantet A., Intermediary object in the context of
distributed design CESA Computational Engineering in systems applications, IEEE-
SMC, Lille, July 96
[11] Fergusson E,S, Engineering and the Mind's Eye MIT press Cambridge, MA 1992
[12] Rodgers PA, Green G., McGownA. Using concept sketches to track design progress in
Design studies 21 N°5 pp 465-481 sept 2000
[13] N. Dodier, les appuis conventionnels de I'action elements de pragmatique sociologique
Reseaux n°62 edition CNET, 1993
[14] Aytes G., Comparing Drawing Tools and Whiteboards: An Analysis of the Group
Process in CSCW 4: 51-71,1996
[15] Laureillard P., conception integree dans I 'usage, PHD thesis INP Grenoble 1999
[16] Boujut, Blanco The role of objects in design co-operation: communication throught
virtual or physical objects. In COOP 2000 conferences Workshop Sophia Antipolis
May 2000
Dr Blanco Eric
Soils Solids Structures laboratory, Domaine Universitaire, BP 53, 38041 GRENOBLE Cedex
9, France, Tel: (33)-4-76-82-70-11, Fax (33)-4-76-82-70-43, mail - eric.blanco@hmg.inpg.fr
Dr Gardoni Mickael
GILCO laboratory, ENSGI, 46 avenue Felix Viallet, 38031 Grenoble Cedexl, France, Tel
(33) (0)4 76 57 43 33, Fax (33) (0)4 76 57 46 95, mail: gardoni@gilco.inpg.fr
(c)IMechE2001
18 ICED 01 - C586/418](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-38-320.jpg)
![INTERNATIONAL CONFERENCE ON ENGINEERING DESIGN
ICED 01 GLASGOW, AUGUST 21-23,2001
AUTOMATIC COMPOSITIONOF XML DOCUMENTSTOEXPRESS
DESIGN INFORMATION NEEDS
A Dong, S Song, J-L Wu, and A MAgogino
Keywords: Information analysis, classification and retrieval; information representation;
design information management
1 Introduction
Engineering design is an information intensive activity. It is reported that designers spend in
excess of 50% of their time in handling information [7]. Thus, the efficiency and the quality
of the design process depend considerably on how well designers are able to handle large
amounts of information. One study [8] of the information required by design engineers to
complete their jobs indicated that less than 50% of that information was actually available and
only 20% could be provided by the existing specialized applications. Directing the right
information to the right person at the right time is a complicated but crucial task.
Design information management has received increasing attention in recent years as a result
of these findings and the recognition that lacking sufficient or missing key design information
may lead to sub-optimal decision-making and design [2,4]. Much of the existing research has
focused on the capture, storage, indexing and presentation of design information including
informal information [3,9]. Less work has been done on information retrieval based on an
understanding of individual designers, their experience, their skills and the ways in which
they use information in the context of their design task [1]. One key step in finding the right
information is expressing information needs in context. This paper presents a methodology to
generate an XML (extensible Markup Language) document that expresses the information
needs of a design engineer. XML documents and their underlying Document Type Definition
(DTD) offer an efficient structure for the organization of design information [5] and
representation of information needs. Through declarations of XML entities and the inherent
structural hierarchy of XML documents, XML documents can express the designer's
information needs while framing the design's structural hierarchies. For example, an element
in the DTD may permit the design engineer to express a preference for formal company
documents of past designs (e.g., technical memos) rather than informal design notes. The
data in the XML document may be drawn from a repository of semi-structured or
unstructured text documents that the design engineer has retrieved and placed into a personal
information store by using an information retrieval system.
Our methodology draws from the computational linguistics techniques of natural language
processing and latent semantic analysis (LSA). We assume there exists some underlying
information needs that are expressed by the type of information the engineering designer
wishes to view and download into a personal information store. The "type of information"
will be distinguished primarily by subject but may include other identifiers such as the format
of the information and the intended audience. By applying these linguistic techniques, we can
ICED 01 - C586/422 19](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-39-320.jpg)
![construct an XML document that is descriptive of this underlying need and contains
information directly from the documents that is consistent with the major patterns of
information preferences of the designer.
2 Methodology
2.1 Technical Approach
The methodology for automated composition of the XML documents proceeds along two
axes: 1) explicitly solicit information needs from the designer through standard information
retrieval means; 2) implicitly monitor the information retrieval behaviour of the designer to
information sources including the type, quality and information contained in the documents
the designer chose to retrieve. We test this methodology on access to unstructured
engineering data, such as full-text, because unstructured, textual documents are the principal
mode of communication by engineers. Before proceeding to a detailed discussion of our
methodology, we discuss two core technologies to the methodology: mining of transaction
logs to learn information needs implicitly and computational linguistics.
2.2 Learning Information Needs Implicitly
Many difficulties exist in determining what information a person wants to see as well as
modelling user information needs. Most information retrieval systems require that people
express information needs through a set of keywords or key phrases. However, ascertaining
information needs simply by a word or two is inadequate and subject to loss of contextual
information. For engineering design, studies have shown varying information needs of
designers depending on level of expertise and stage of the design [1,6]. Our approach in
modelling user information needs is based on a human-centred computing. Our methodology
examines the user's document access patterns, that is, the user's personal information store
and transaction history in an information retrieval system, for patterns of information
preference. The basic approach is to examine the user's session information over all sessions
while using the information management system. In learning information needs implicitly,
we are primarily interested in discovering the similarity between documents that the user has
downloaded into a personal information store. The assumption is that the user's particular
choices of documents to store locally are indicative of information needs. Thus, instead of
requiring the designer to a priori categorize the information, the system attempts to learn a
similarity mapping using contextual clues such as project name, engineering discipline, and
document format. Similar categorisation strategies have also been found in the classification
of supplier information practised in industry [12].
To ascertain relevance, the system records each document downloaded into the user's
personal information store. This is an accurate indicator of relevance because, in our system,
before the user may download the document, the user has already read a brief description of
the content of the document containing meta-information such as abstract, author, document
type, and subject. The assumption is that during any single session utilizing the information
retrieval system, the user has a dominant goal (information need) in mind that is expressed by
the type(s) of documents the user decides to download. Similar assumptions exist in other
studies [13] of information retrieval systems.
We use the vector space approach [14] for document and query representation. We
analytically modelled information needs as a linear combination of the vectors representing
20 ICED 01 - C586/422](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-40-320.jpg)
![the query and the relevant document. The weighted vector average (arithmetic mean) of all
combined vectors consisting of the query and relevant documents is called the "centroid" of
the user's intended information needs. The centroid is then used as a representation of the
dominant information need of the user. The maximum angle between the document vectors
or query string vector and the centroid is used as an indication of the variation in the user's
information needs.
2.3 ComputationalLinguisticApproaches
Our methodology employs two computational linguistics techniques, natural language
processing and latent semantic analysis, to extract and summarizethe primary topic of a set of
similar documents.
Natural Language Processing
While we use the designer's past transactions as an indication of information need, the
structured information contained in the documents that are referenced in the transactions
needs to be discovered and refined before it can be utilised effectively. The key phrase
retrieval process helps in crosschecking the indexed subjects and compacting the size of the
aggregated XML document by representing paragraphs of text with just a few representative
noun phrases. As the designer adds documents to a personal information store, we do not
need to concatenate all the textual information about the document to the XML document, just
the extracted noun phrases that are not already in there and plug them under the appropriate
tags. By identifying the noun phrases in the document, we are able to find the corresponding
contents for each XML tag in the text. Additional rules and procedures are needed other than
standard noun phrase retrieval process to perform this task for all tags. This latter component
forms a part of future research.
Extracting from the full-text content-bearing noun phrases documents that can be used in
profiling and indexing involves 3 steps: tokenization, part-of-speech (POS) tagging and noun-
phrase identification. Tokenization is a procedure that identifies sentence boundaries and
removes extraneous punctuations. POS taggers then take the processed corpus and tag each
word with their POS information. Taggers that operate following semantic rules or just
statistical information were developed. After the text corpus has been tagged with POS
information, we could use the contextual information to identify noun phrases. The extracted
noun phrases are then attached to the corresponding DTD elements of the document.
Latent Semantic Analysis
Latent Semantic Analysis (LSA) [9] is a statistical model of word usage that permits
comparisons of semantic similarity between pieces of textual information. The idea is that the
totality of information about all the word contexts in which a word does and does not appear
provides a set of mutual constraints that largely determines the similarity of meaning of words
and sets of words to each other. The primary assumption of LSA is that there exists an
underlying or "latent" structure in the pattern of word usage across documents. LSA uses the
matrix technique of singular value decomposition (SVD) to reflect the major associative
patterns of words in the document and to ignore the smaller influences. The ability for LSA
to remove the obscuring "noise" makes LSA useful as an analytical tool for discovering the
primary conceptual content of documents. We use LSA to help us to categorize and group the
documents by topic material that the user has downloaded and revealed as relevant and useful.
Once these principal groupings are identified, we can then apply the rest of our methodology
ICED 01 - C586/422 21](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-41-320.jpg)
![to express information needs for each "centroid" of documents within a semantic locality
through a single XML document.
2.4 ComposingXML Documents
The system initiates by asking the user to specify an XML DTD containing metadata elements
(XML entities) that express information needs. For this study, we asked that users express
their information needs using a well-known metadata set, the IEEE Learning Object Metadata
(LOM) [10], and in particular the core 20 elements. The LOM defines the information
required to manage, locate and evaluate learning resources. Having a standards-based XML
DTD allows us to assess our methodology's completeness and efficacy and for comparison
against other systems. Because of the analogous human cognitive processes of information
processing and learning [11], the LOM serves as a reasonable model for describing
information needs.
Once the user has specified an XML DTD, the user must now utilise the information
management system, retrieving and downloading design documents. The system implicitly
monitors the information retrieval transactions of the user, eventually formulating a seed an
XML document containing information from the user's session. Typically, the XML
document is seeded with the initial query the user posed to the system.
The implicit stage contains two phases. In phase one, the system applies latent semantic
analysis to characterize the knowledge conveyed by the all documents the person chose to
view. To perform this phase, we analysed the transaction logs containing the transactions of
both the current user and all other users of the system. The user's query and document(s)
downloaded are recorded for each visit. The latter information identifies the relevant
documents necessary for phase two. Using a similarity measurement, the system identifies
topics represented by the documents the user viewed. By doing this step, the system
ascertains the topic locality of the various documents the person viewed. This is a critical step
because the user may have multiple and widely varying information needs. Then, for each
topic locality, the system augments the XML document expressing the user's information
needs with the metadata elements from each relevant document. In practice, the information
management system will contain most of the information required to complete the tagging
such as Author, Title, Date, and Format. Subject and Description information are generated
automatically from the second phase. This matching can be done by exact one-to-one
correlation, i.e., both the tag and attribute match, or via a crosswalk between the information
about the document contained in the information management system and the XML DTD.
Both of these techniques will have required some prior means of tagging the documents in the
information management system.
In the second phase, we apply natural language processing techniques to ascertain the
principal subject of the documents within each topic as discussed in Section 2.3. The process
repeats until all possible elements in the user's original XML DTD are filled, resulting in a
fully marked up XML document. The completed DTD is now an expression of the
information needs of the designer, based upon the available information stores and the pattern
of information retrieval undertaken.
In summary, a designer's information needs can be found implicitly by looking at the set of
documents the designer has deemed relevant and useful and stored in a personal information
store. We use LSA to find signatures of similarity in this set of documentation. Once we find
the signatures, we look at what the original information needs were to find the centroid of the
22 ICED01 - C586/422](https://image.slidesharecdn.com/42530-250223200335-4e6eccc2/85/Design-Management-Process-and-Information-Issues-Iced-Issues-v-1st-Edition-S-Culley-42-320.jpg)
Design Management Process and Information Issues Iced Issues v 1st Edition S. Culley
- 1. Visit https://ebookfinal.com to download the full version and explore more ebooks Design Management Process and Information Issues Iced Issues v 1st Edition S. Culley _____ Click the link below to download _____ https://ebookfinal.com/download/design-management- process-and-information-issues-iced-issues-v-1st- edition-s-culley/ Explore and download more ebooks at ebookfinal.com
- 2. Here are some suggested products you might be interested in. Click the link to download The Fourth Amendment Select Issues and Cases Select Issues and Cases 1st Edition Jesse V. Kessler https://ebookfinal.com/download/the-fourth-amendment-select-issues- and-cases-select-issues-and-cases-1st-edition-jesse-v-kessler/ Ergonomics in the Automotive Design Process Concepts Issues and Methods second edition Vivek D. Bhise https://ebookfinal.com/download/ergonomics-in-the-automotive-design- process-concepts-issues-and-methods-second-edition-vivek-d-bhise/ Same Sex Issues 1st Edition Jason V. Albertson https://ebookfinal.com/download/same-sex-issues-1st-edition-jason-v- albertson/ Pipelines for Carbon Sequestration Background and Issues Background and Issues 1st Edition Elvira S. Hoffmann https://ebookfinal.com/download/pipelines-for-carbon-sequestration- background-and-issues-background-and-issues-1st-edition-elvira-s- hoffmann/
- 3. Global Issues in Food Science and Technology 1st Edition Gustavo V. Barbosa-Canovas https://ebookfinal.com/download/global-issues-in-food-science-and- technology-1st-edition-gustavo-v-barbosa-canovas/ Strategic Management Issues and Cases 2nd Edition Paul W. Dobson https://ebookfinal.com/download/strategic-management-issues-and- cases-2nd-edition-paul-w-dobson/ Mexico Background and Issues Background and Issues 1st Edition Richard C. Bradley https://ebookfinal.com/download/mexico-background-and-issues- background-and-issues-1st-edition-richard-c-bradley/ Information Security Cost Management 1st Edition Ioana V. Bazavan https://ebookfinal.com/download/information-security-cost- management-1st-edition-ioana-v-bazavan/ Early Childhood Education Issues and Developments Issues and Developments 1st Edition Petr G. Grotewell https://ebookfinal.com/download/early-childhood-education-issues-and- developments-issues-and-developments-1st-edition-petr-g-grotewell/
- 5. Design Management Process and Information Issues Iced Issues v 1st Edition S. Culley Digital Instant Download Author(s): S. Culley ISBN(s): 9781860583551, 1860583555 Edition: 1 File Details: PDF, 39.42 MB Year: 2001 Language: english
- 7. Design Management - Process andInformation Issues
- 8. WDK Publications WDKl WDK 2a WDK 2b WDK 3 WDK 4a WDK4b WDK 4c WDK 5 WDK 6 WDK 7 WDK 8 WDK 9 WDK 10 WDK 11 WDK 12 WDK 13 WDK 14 WDK 15 WDK 16 WDK 17 WDK 18 WDK 19 WDK 20 WDK 21 WDK 22 WDK 23 WDK 24 WDK 25 WDK 26 WDK 27 WDK 28 Principles of Engineering Design Bibliography of Design Science Bibliography of Design Science (continued) Terminology of the Science of Design Engineering in Six languages Case Examples(1-3) Case Examples (4-6) Case Examples (7-9) Design Methodology: Proceedings ICED 81, Rome Teaching Engineering: Proceedings ICED 81, Rome Conference Results - ICED 81,Rome J Dietrych about Engineering Design Tactics in Engineering Design (Readings) Proceedings ICED 83, Copenhagen Management of Engineering Design (Readings) Proceedings ICED 85, Hamburg Proceedings ICED 87, Boston Methodical Design of Machine Elements (Readings) Reliability of Technical Systems Proceedings ICED 88, Budapest EVAD - Evaluation andDecision inDesign (Readings) Proceedings ICED 89, Harrogate Proceedings ICED 90, Dubrovnik Proceedings ICED 91, Zurich Engineering Design Education (Readings) Proceedings ICED 93, The Hague Proceedings ICED 95, Prague EDC - Engineering Design andCreativity, Workshop Proceedings Proceedings ICED 97, Tampere Proceedings ICED 99, Munich Manual for Design Engineering (Selected preprint) Proceedings ICED 01, Glasgow
- 9. 13th International Conference on Engineering Design - ICED 01 Design Management - Process and Information Issues 21-23 August 2001 Scottish Exhibition and Conference Centre, Glasgow, UK Organized by The Institution of Mechanical Engineers (IMechE) Sponsored by BAE SYSTEMS Co-sponsored by The Institute of Engineering Designers The American Society of Mechanical Engineering (ASME) Published by Professional Engineering Publishing Limited for The Institution of Mechanical Engineers, Bury St Edmunds and London, UK.
- 10. First Published 2001 This publication is copyright under the Berne Convention and the International Copyright Convention. AH rights reserved. Apart from any fair dealing for the purpose of private study, research, criticism or review, as permitted under the Copyright, Designs and Patents Act, 1988, no part may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, electrical, chemical, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owners. Unlicensed multiple copying of the contents of this publication is illegal. Inquiries should be addressed to: The Publishing Editor, Professional Engineering Publishing Limited, Northgate Avenue, Bury St Edmunds, Suffolk, IP32 6BW, UK. Fax: +44 (0) 1284 705271. © 2001 The Institution of Mechanical Engineers, unless otherwise stated. ISBN 1 86058 355 5 A CIP catalogue record for this book is available from the British Library. Printed by The Cromwell Press, Trowbridge, Wiltshire, UK The Publishers are not responsible for any statement made in this publication. Data, discussion, and conclusions developed by authors are for information only and are not intended for use without independent substantiating investigation on the part of potential users. Opinions expressed are those of the Author and are not necessarily those of the Institutionof Mechanical Engineers or its Publishers.
- 11. Conference Organizing Team Steve Culley, Bath University,UK Alex Duffy, Strathclyde University,UK Chris McMahon, Bristol University,UK Ken Wallace, Cambridge University, UK Scientific Advisory Board A Chakrabarti, University of Cambridge, UK R Anderl, Technische Universitat Darmstadt, Germany R Andrade, Universidade Federal do Rio de Janiero, Brasil M M Andreasen, Technical University of Denmark, Denmark E K Antonsson, Caltech, USA P Badke-Schaub, Universitat Bamberg, Germany A H Basson, University of Stellenbosch, South Africa H Birkhofer, Technische Universitat Darmstadt, Germany L T M Blessing, University of Cambridge, UK G Blount, Coventry University, UK A Breiing, Swiss Federal Institute of Technology, Switzerland S Burgess, University of Bristol, UK J Busby, Cranfield University, UK M Cantamessa, Politecnico di Torino, Italy H H C M Christiaans, Delft University of Technology, The Netherlands P J Clarkson, Universityof Cambridge, UK J Deans, University of Auckland, New Zealand P Deasley, Cranfield University, UK W E Eder, Royal Military College of Canada, Canada K Ehrlenspiel, Technische Universitat Munchen, German) S D Eppinger, Massachusetts Institute of Technology, USA D G Feldmann, Technische Universitat Hamburg- Harburg, Germany S Finger, Carnegie Mellon University, USA P Frise, University of Windsor, Canada I Gibson, National University of Ireland, Ireland H Grabowski, Universitat Karlsruhe, Germany G Green, University of Galsgow, UK K-H Grote, Otto-von-Guericke-Universitat Magdeburg, Germany P H K Hansen, Aalborg University, Denmark I Horvath, Delft University of Technology, The Netherlands St Hosnedl, University of West Bohemia in Pilsen, Czech Republic V Hubka, Heurista, Switzerland B Ion, University of Strathclyde, UK D Jiati, 720 Lab, Bejing University of Aeronautics and Astronautics, Peoples' Republic of China N Juster, University of Strathclyde, B Katalinic, TU Vienna, Austria T Kiriyama, Stanford University, USA S S Kok, Nanyang Technological University, Singapore L Leifer, Stanford Centre for Design Research, USA B Lewis, University of Melbourne, Australia Yn Li, Sichuan University, Peoples' of Republic China Uo Lindemann, Technische Universitat Munchen, Germany M L Maher, University of Sydney, Australia M Mantyla, Helsinki University of Technology, Finland D Marjanovic, University of Zagreb, Croatia H Meerkamm, UniversitatErlangen-Nurnberg, Germany M Meier, ETH-Zurich, Switzerland F Mistree, Georgia Institute of Technology, USA M Norell, Royal Institute of Technology, Sweden M Ognjanovic, University of Belgrade, Yugoslavia K Otto, Massachusetts Institute of Technology, USA U Pighini, University of Rome 'La Sapienza', Italy D F Radcliffe, The University of Queensland, Australia Y Reich, Tel Aviv University, Israel A Riitahuhta, Tampere University of Technology, Finland R Rohatynski, Technical University of Zielona Gora, Poland N Roozenburg, Delft University of Technology, The Netherlands E Rovida, Politecnico di Milano, Italy A Samuel, University of Melbourne, Australia J W Schregenberger, ETH Honggerberg, Switzerland P Sen, University of Newcastle, UK M Simon, Sheffield Hallam University, UK J Simmons, Heriot-Watt University, UK L Stauffer, University of Idaho, USA K G Swift, University of Hull, UK O Tegel, TU Berlin, Germany G Thompson, UMIST, UK D L Thurston, University of Illinois at Urbana- Champaign, USA M Tollenaere, ENS Genie Industriel, France T Tomiyama, The University of Tokyo, Japan D G Ullman, Oregon State University, USA S Vajna, Otto-von-Guericke-Universitat Magdeburg, Germany C Weber, Universitat des Saarlandes, Germany M P Weiss, TECHN1ON, Israel P M Wognum, University of Twente, The Netherlands K Wood, The University of Texas, USA M M F Yuen, Hong Kong University of Science and Technology, Peoples' Republic of China
- 12. Industrial Advisory Board R Bodington, BAE Systems Advanced Technology Centre, UK D Knott, Rolls-Royce plc, UK N Brenchley, Forward Industries, UK I Liddell, Buro Happold, UK M Brown, BAE Systems Research Centre, UK P W Liddell, Lockheed/BAe, USA P Burnell, EPSRC, UK K Mashford, Interfacing Limited, UK I Chatting, GKN Westland Aerospace, UK I Milbum, Nissan European Technology Centre Limited, UK A Court, Portacel, UK J Mudway, TRW ASG Lucas Aerospace, UK P Fearis, The Generics Group Limited, UK Y Neuvo, Nokia Group, Finland P Ferreirinha, MIRAKON, Switzerland F O'Donnell, Scottish Enterprise Lanarkshire, UK D Foxley, Royal Academy of Engineering, UK C Pearce, INBIS Group plc, UK E Frankenberger, Heidelberger Druckmaschinen AG, B Prasad, CERA Institute, USA Germany D Rimmer, Pilkington Optronics Limited, Denbighshire, UK N Grant, BAE Systems Operations, UK D Robson, Scottish Enterprise, UK J Gunther, Hilti AG, Liechtenstein M Shears, Ove Arup Partnership, UK C Hales, Triodyne Inc., USA L Styger, STYLES RPD, UK L Hein, Technical University of Denmark, Denmark I Yates, UK N Kohlhase, LEWA Herbert Ott GmbH + Co., Germany
- 13. Contents Preface Knowledge and Information Management Information Management Improving access to design solution spaces using visualization and data reduction techniques P M Langdon and A Chakrabarti Supporting non-structured graphical information in integrated design team E Blanco and M Gardoni Automatic composition of XML documents to express design information needs A Dong, S Song, J-L Wu, and A M Agogino Design communication using a variation of the design structure matrix J C Lockledge and F A Salustri Multi - a tool and a method to support collaborative functional design S Menand and M Tollenaere Information flow in engineering companies - problems andtheir causes C M Eckert, P J Clarkson, and M K Stacey Design case relevance in quantity dimension space for case-based design aid T Murakami, J Shimamura, and N Nakajima A web-based information tool for application engineering J Schmidt and D G Feldmann Knowledge Representation andManagement An agent environment to support co-ordination between design actors P Girard and C Merlo Developing a support system for novice designers in industry S Ahmed and K M Wallace Using domain knowledge to support design requirements elicitation M J Darlington, S J Culley, and S Potter Towards pragmatic approaches for knowledge management in engineering - theory and industrial applications K-D Thoben, F Weber, and M Wunram Effective abstraction of engineering knowledge for KBE implementation P Bermell-Garcia, I-S Fan, G Li, R Porter, and D Butter 3 3 11 19 27 35 43 51 59 67 67 75 83 91 99 xii
- 14. New techniques for design knowledge exploration - a comparison of three data grouping approaches P C Matthews, P M Langdon, and K M Wallace Computer-supported systematic design and knowledge in the early design phase E Frankenberger DEKAS - an evolutionary case-based reasoning system to protection scheme design G West, S Strachan, J McDonald, A H B Duffy, J Farrell, and B Gwyn Knowledge for product configuration K Hadj Hamou, E Caillaud, J Lamothe, and M Aldanondo An operational model for design processes M Y Ivashkov and C W A M van Overveld Classifications and Taxonomies A common language for engineering design practice and research A Samuel, J Weir, and W Lewis Using situation theory to model information flow in design Z Wu and A H B Duffy Shape matching and clustering S Lim, A H B Duffy, and B S Lee The product develoment process ontology —creating a learning research community P K Hansen, A Mabogunje, O Eris, and L Leifer The application of an automatic document classification system to aid the organizers of ICED 2001 A Lowe, C A McMahon, T Shah, and S J Culley Analysis and classification methodology for objectifications in collective design processes A Verbeck and K Lauche Design Reuse Modularity in support of design for re-use J S Smith and A H B Duffy Capturing and classifying information in undergraduate design team projects P A Rodgers Incorporating incentives into design documentation tools T Lloyd and L Leifer 107 115 123 131 139 147 147 155 163 171 179 187 195 195 203 211
- 15. Scaling-up domain knowledge representation in the development of knowledge intensive CAD for proactive design for manufacture X-T Yan, J Borg, and F Rehman Re-using knowledge - why, what, andwhere J S Smith and A H B Duffy Documentation and evaluation in early stages of the product development process L Schwankl Mapping experience - learning from experience of peers through socio- technical interactions T Liang, D G Bell, and L J Leifer Transfer of experience in critical design situations P Badke-Schaub, J Stempfle, and S Wallmeier Organization and Management of Design Project Management Product development models for shorter lead times and more rational processes - its effects on the work situation for project managers and project group A Zika-Viktorsson and J Pilemalm A multi-project management approach for increased planning process F Marie and J-C Bocquet Applying conceptual design methods to engineering management P Hughes, C Burvill, and R Hughes Finding and implementing best practices for design research activities M Gardoni and C Frank Project Strategy and Management Late point differentiation analysis for configurable products T A Lehtonen, A O Riitahuhta, and J M Malvisalo On the design of services R Andrade Functional products create new demands on product development organizations O Brannstrom, B-O Elstrom, and G Thompson How missions determine the characteristics of product development methodologies B Bender, B Bender, and L T M Blessing 219 227 235 243 251 261 261 269 277 285 293 293 299 305 313
- 16. Planning and Workflow Management Visualization techniques to assist design process planning P J Clarkson, A F Melo, and C M Eckert A product and process model supporting main and sub-supplier collaboration B Fagerstrom and H Johannesson Identifying and analysing changes in a dynamic company S Suistoranta Design process planning using a state-action model A F Melo and P J Clarkson Concurrent Engineering and Integrated Product Development Integrated newproduct development - a case-based approach R Valkenburg and J Buijs Material instrumentation forinter-trade co-operation - a source of innovation: application in the domain of painting and varnish finishes N Stoeltzlen, D Millet, and A Aoussat Geometry users from a process perspective F Fuxin Concurrent design of product and package- extending the concept of IPD C Bramklev, R Bjarnemo, and G Jonson Distributed Design/Supply Chain Integration A framework for distributed conceptual design A Schueller and A H Basson A system for co-ordinating concurrent engineering R I Whitfield, G Coates, A H B Duffy, and W Hills Distributed product development - a case study in inter-organizational SME business networks J Pilemalm, S Gullander, P Norling, and A Ohrvall-Ronnback Organizational design - a tool for evaluating alternative extended enterprise structures A McKay and A de Pennington Design Teams Cultural issues in aerospace engineering design teams K H Payne and P J Deasley Supporting the teamwork by new means of the information technology A M Kunz, S Muller, T Kennel, K Lauche, and K Mbiti The importance of informal networks to effective design management N J Brookes, P Smart, and F Lettice 321 321 329 337 345 353 353 361 369 377 385 385 393 401 409 417 417 425 433
- 17. Managing uncertainty in design communication M K Stacey and C M Eckert Managing the integration between design, research, and production in the automobile industry H V de Medina and R M Naveiro Researching the thinking process in design teams - an analysis of team communciation J Stempfle and P Badke-Schaub Towards a science of engineering design teams A Mabogunje, K Carrizosa, S Sheppard, and L Leifer Dimensions of communication in design C M Eckert and M K Stacey A statistical study of how differing levels of diversity affect the performance of design teams A G Carrillo Management of the Clarification Phase Requirements engineering - laying thefoundations for successful design G A Thomson The organization and management of engineering tenders G Barr, J H Sims Williams, S C Burgess, and P J Clarkson Modification of a methodological design tool for the developing country scenario - a case study in product definition K M Donaldson and S D Sheppard An approach for structuring design specifications for complex systems by optimization C Grante, M Williander, P Krus, and J-O Palmberg An engineering approach for matching technology to product applications J B Larsen, S P Magleby, and L L Howell Performance Evaluation Financing innovation in co-operation projects P Link and S Spiroudis Performance management at design activity level F J O'Donnell and A H B Duffy A metrics methodology developed in co-operation with industry M W Lindley, M Muranami, and D G Ullman Improvement of engineering processes S Vajna, D Freisleben, and M Schabacker 441 449 457 465 473 481 489 497 505 513 521 529 529 537 545 553
- 18. Process performance measurement support - a critical analysis M K D Haffey and A H B Duffy Risk and Uncertainty Management The methodology for system integrity in design J K Raine, D Pons, and K Whybrew Change prediction for product redesign P J Clarkson, C Simons, and C M Eckert Survey of current UK practice in managing technical design risk R Crossland, C A McMahon, and J H Sims Williams Development of an 'IDEA' for safety D Vassalos, I Oestvik, and D Konovessis How mutual misconceptions between designers and operators cause accidents in hazardous installations J S Busby, R E Hibberd, P W H Chung, B P Das, and E J Hughes A project view of the handling of uncertainties in complex product development organizations R Olsson Decision-making - howto avoid dysfunctions? Howto analyse dysfunctions? How to improve an organization by its dysfunctions? J Stal-le Cardinal, M Mekhilef, and J-C Bocquet Preliminary design of a risk management decision tool J M Feland Authors' Index 561 569 569 577 585 593 601 609 617 625 633
- 19. Preface This is one of four books resulting from the contributions to the 13th International Conference on Engineering Design (ICED 01). The conference was held in August 2001 in the Scottish Exhibition and Conference Centre located on the River Clyde in Glasgow - the ideal place to hold the first ICED of the nowmillennium. The ICED conference series was initiated by Workshop Dcsign-Konstruktion (WDK) in 1981 with the first conference in Rome. From the very beginning, the aim of ICED was to offer a platform for the discussion of new trends, developments, and research findings in the areas of new product development, design support techniques, design processes, design science, and design education. The conferences have been held in eleven different countries and have become one of the most pre-eminent conferences in the field of Engineering Design, with the last two conferences being held in Tampere, Finland, in 1997 and in Munich, Germany, in 1999. Both these conferences attracted well over 500 delegates from both academic institutions and industrial organizations. Nearly all the leading authorities in the field of Engineering Design attend to report their latest findings and exchange current ideas with colleagues. All conferences have focussed on the process of planning, developing and designing technical systems and products. ICED covers all aspects and disciplines of engineering design, from general product development and innovation to feature-based geometric reasoning and design for later life-phases. As engineering design is a process to which many disciplines are contributing, an additional emphasis has been placed on design management, organization, teams, and individuals. Over the years ICED conferences have become the forum for establishing, maintaining, and improving contacts and co-operation between researchers and engineers from countries all over the world. It is self evident that the engineering design process has changed to meet the challenges of globalization, increasing international competition, and the need for sustainable development. Equally the performance and quality of engineering products have improved in many aspects, time to market, performance, reliability, reduced environmental impact, etc. If the improvements are to be maintained, the elements that contribute to the product development process must continue to be studied and enhanced. Improvements in the engineering design and its process have been supported by theories and methods developed by research groups around the world. The research is beginning to mature into an overall and consistent understanding of engineering design as will be seen in the pages of the four books. However the results are still fragmented and there is a need to unify the findings, and to ensure that these findings are transferred into industry. The theme chosen for ICED 01 was Unifying Engineering Design - Building a Partnership between Research and Industry. The organizing team received 664 Abstracts and this resulted in some 325 full papers. All papers are eight pages in length and went through a double blind review of the abstracts and a double review of the full papers. The books consist of contribution papers from some 35 countries. xiii
- 20. DESIGN MANAGEMENT This book consists of some 13 topics and really has an overarching theme of the management of the process and the information that supports the process. It covers knowledge and information management at all levels and the organization and management issues associated with the design activity itself. Books in the series Book 1 Design Research Book 2 Design Management Book 3 Design Methods Book 4 Design Applications A large number of people and organizations have helped with the conference. The organizing team would like to express their thanks to all who have contributed to the content and execution of ICED01 in whatever way. Steve Culley Alex Duffy Chris McMahon Ken Wallace
- 21. Knowledge and Information Management Including sub-sections: Information Management Knowledge Representation and Management Classifications and Taxonomies Design Reuse
- 22. This page intentionally left blank
- 23. INTERNATIONAL CONFERENCE ON ENGINEERING DESIGN ICED 01 GLASGOW.AUGUST21-23, 2001 IMPROVING ACCESS TO DESIGN SOLUTION SPACES USING VISUALIZATION AND DATA REDUCTION TECHNIQUES P M Langdon and AChakrabarti Keywords: Visualisation, clustering, synthesis, design information management, information analysis, classification and retrieval 1 Introduction Designers only explore a few solutions in depth at the conceptual stage [1]. Despite this, evidence suggests that a thorough exploration of a solution space is more likely to lead to designs of higher quality [2]. FuncSION [3] is a computational tool that synthesises a wide range of solutions to a class of mechanical design problems involving transmission and transformation of mechanical forces and motions that are specified as inputs and outputs. It uses a set of primary functional elements along with combination rules to create anexhaustive set of solutions in terms of their topological and spatial configurations. Previous research has demonstrated FuncSION's potential for generating novel solution ideas that designers had not thought of. However, it was found that the large number of ideas generated could not be meaningfully explored, while their representation proved too abstract to visualise. Effective support for conceptual design should help designers obtain a thorough overview of the solution space, as well as a detailed understanding of its individual solutions. However, the greater the variety and number of solutions to be explored, the less likely it is that a detailed understanding of the potential of all individual solutions will be achieved. The DESYN software described in this paper encapsulates FuncSION with a Graphic User Interfacce (GUI) [4]. The overall aim of this work is to test the extent to which DESYN is capable of assisting designers in their exploration of large solution spaces such that an overview of the entire space is obtained while at the same time facilitating understanding of the individual solutions contained in it. 2 Background A paper presented at ICED99 [4] described a scheme adopted in order to solve the above problems. The problem of exploring large combinatorial spaces is an unresolved issue in computer aided synthesis research, which is further complicated by the difficulty in displaying large volume of information [5]. The approach adopted was a novel method of clusteringthe solution configurationsto reduce the space of solutions that the designer needs to consider in order to get an overview of the entire space of solutions [6]. This was done by presenting the designers with representative solutions that were by-products of the clustering process. In this way, large numbers of solutions can be summarised by a small number of cluster exemplars of prototypes. ICED 01 - C586/223 3
- 24. A number of previous approaches have adopted rule-based heuristics for the generation of solution compositions in an unconstrained space (See, for example, Campbell et al, 1999)[7] [8][9]. However, the approach adopted here assumes exhaustive generation of solutions in a partially constrained space with the use of clustering as a data reduction and representation. Because of this, all solutions are, in principle, available to the designer through interaction with the systems GUI. In the authors' earlier paper, some initial validation of the cluster algorithm was reported focussing on whether the system's clusterings of solutions were intuitive and whether the clusterings suggested by the system correspond to designers' own partition of the solution space. Initial experiments examined a number of DESYN clusterings of two sets of 20 solutions resulting from a synthis using 0-5 elements per solution from a choice of < 4 elements. Figure 1 is an example diagram of the outputs of the same clustering algorithm for a smaller 6 solution set synthesised using the elements wedge, cam and lever mechanisms. Clustering was based on a Euclidian distance metric operating on a count of features in each solution from a feature set of all possible feature pairs. Figure 1.Cluster table for 2 to 5 clusteringsof a 6 solutions for a problem using Wedge, Cam and Lever elements. The arrows indicate the solutions that leave the 3 cluster solution to form new clusters. On the left of the diagram the six possible solutions output by the synthesiser are enumerated. The vertical columns show increasing number of clusters in the solution set. The cells show the cluster membership for each solution. Each cluster has a representative solution that is denoted by circling. Finally,the box in the corner of cells shows the average distance of the representative to all the other members of the cluster. The right hand diagram shows a Venn type representation of the 3,4 and 5 cluster solutions. Solution sets for the larger 5 element problems were presented to 5 subjects as printed words with illustrative diagrams. Subjects were required to form groupings of the 20 solutions that corresponded to their judgements of those that seemed to them to be similar on the basis of their engineering experience. The subject's groupings were scored for the percentage similarity of groupings with the DESYN clusterings of the same set. This was to test whether designers found the system's clustering of solutions intuitive, and whether the clustering suggested by the designers correspond to their own partition of the solution space. This clustering was based on an element-pair feature count (focusing on the type of interfaces possible as design elements in combination),and provided an average commonality, between designer's clustering and the 4 ICED 01 - C586/223
- 25. method's, of 74% in contrast to 55% commonality yielded by comparison with a random clustering. However, this evaluation was carried out using only the topology, rather than the 3-D configurations of the solutions. Evaluation of the approach is further extended in this paper (see section 3.2) using: (1) 3-D configurations displayed by the system as displays of force and motion solution chains; (2) Graphic representations of the clustered solution space. The goal of this evaluation is to find whether or not this clustering-based technique improves a designer's overview of the entire solution space, and whether the software facilities assist their understanding of the solution space. 3 Current Developments Current developments include further development of the GUI and visualisation support, and further evaluation of DESYN, which are described below. 3.1 User interface development The visualisation utilises a 2D 'star' representation of clusters (Figure 2), and a pseudo-3D display of a component chain schematic (Figure 3) that we have developed to improve access to the functional synthesis software. The former is used to aid visualisation of the spatial relationships between the component interfaces, while the latter is intended to aid the understanding of the grouping of functional solutions by similarity [10, 11]. Figure 2. The DESYN Graphic User Interface Clustering Tool In conjunction with this, a 3D visualisation for displaying an overview of the solution space and the spatial layouts of selected solutions resulting from the synthesis is under development. ICED 01 - C586/223 5
- 26. When implemented, this will display a 3D schematic representation of individual solution chains with a symbolic convention used to indicate locations and directions of force and rotation. Elements will represented by shaded cylinders and interfaces between elements are represented by spheres. Figure 3. The Pseudo 3D visualisation interface The orientation, direction, and sense of forces will be represented by arrow triplets. The Nominal direction of rotation or force was represented by 2D sprites. Elements will be fitted into the mechanism boundaries by an algorithm that distributes the element lengths evenly into the available space. Figure 3 represents a schematic of a concept that has four elements, nominally oriented as shown by the arrows, such that an input rotation is taken by the first (shaft-like) element, and passed on to the second (crank-like) element, which passes the resulting translation via two tie rod-like elements to the output point. The 3D representation and facilities for manipulationof individual solutions are intended to provide designers a more detailed understanding of the individual solutions than that provided by the more abstract text-based representation. 3.2 Evaluation Each of two groups of designers (Gl and G2) were asked to individuallyinspect the solutions set generated by FuncSION in two conditions by using the softwares Graphic User Interface. The unstructured group were presented with a list representation of the solution set and the structured group received the representation in the form of 2D "sun-and-planet" graphic representing clusters calculated by the algorithm, the representative central member and the average distance of cluster members from that medoid (Figure 2). The task required the designers to select, using the representation available to them, a set of solutions to the specified problem that were both different from each other and formed a set completely representative of all solutions. In the unstructured condition the designers were given a text field on the interface into which they could fetch groups of solution on demand. This was managed by the use of a button that simply obtained five solutions from the total set and placed them at the bottom of the list on 6 1CED 01 - C586/223
- 27. the screen. No indication of the total number of solutions was given and the solutions were not presented in any ordering linked to their functional structure. Hence, the designers were allowed to select as few or as many solutions as they liked. The task further required that the designers selected individual solutions form the list (left, Figure 4) that they felt met the criteria. This was made on a point-and-click basis and the resulting set stored in a further text field (right, Figure 4). Figure 4. The list representation experimentalgroup interface. In the structured condition the designers performed the same task but interacted with a 2D cluster representation of the solutions set. Dwelling the cursor on a central representative "sun" member of a cluster lead to a list of the cluster solution members appearing in the upper right text field. Dwelling on the numbered "planet" members highlighted the location of the solution in the list. Figure 5. The list representation experimentalgroup interface. Again, the task further required that the designers select individual solutions form the list (top right, Figure 5) that they felt met the criteria. This was made on a point-and-click basis by clicking on the member "planets" and the resulting set stored in a further text field (bottom right, Figure 5). In both conditions the designers were able to delete members of their selected set at will. 3.3 Data analysis technique Results in our previous study [4] suggested that the clustering solution for small problems had some psychological validity as a grouping of the solution space correlated highly with designers groupings. Therefore, in the present case, if a solution chosen by a designer belonged to a cluster generated by the computer, then it was assumed that the designer had an overview of at least that cluster. If there was a better match between the computed clusters ICED 01 - C586/223 7
- 28. and solutions listed by the designers in the list condition (Gl) rather than with those listed by the designers in the clustered condition (G2),then it could be concluded that designers had a better overview of the solution space using clustering than without. This would indicate that a designer could miss innovative solutions when they did not use the data reduction technique. For the list of concepts written by each designer, solution clusters were generated using the clustering algorithm, taking the size of the list as the number of clusters. A comparison was then made between the list of concepts of the designer with the solution clusters generated by the algorithm to find how many solutions from the designer's list belong to distinct computed clusters. The proportion of the solution space covered by the designer, taken as a measure of the overview obtained by him, was then calculated as the ratio of the number of clusters 'covered' by the designer to the total number of clusters. An average proportion of solution space covered by designers in each group was then calculated by dividing the total of the ratios (for all the designers in the group) by the number of designers. 4 Results List Condition (G1) Subject S1 S2 S3 Total Average No. 3 3 7 13 4.3 Matched 1 1 6 8 2.7 Ratio 1/3 1/3 6/7 21/32 0.51 Clustered Condition (G2) Subject S4 S5 S6 Total Average No. 3 3 7 11 3.7 Matched 1 1 5 9 3.0 Ratio 1/3 1/3 5/7 21/29 0.46 Table 1. Cluster commonality data for the list conditionand clustered conditions The results obtained are tabulated in Table 1. Very little difference was obtained between the numbers of clusters sampled by the designers in each experimental group and this was reflected by the ratios of computed and designer sampled clusters (Table 1.). Although no inferential statistics were calculated due to the small size of the sample (n = 3 in each group) it is evident from the table that many of the designers in both groups chose a small set of three solutions that sampled only one cluster of the computed solutions. 5 Discussion There was no evidence that the designers sampled the solution space more effectively using the 2D cluster visualisation GUI when measured in terms of the similarity between the clusters they sampled and the complete optimal clusterings that had been previously found to correspond to designers' intuitive groupings. The principal result suggests that the use of an assistive aid combining data reduction technique with graphic visualisation did not lead to a more complete sampling of the solution space for the highly constrained synthesis problems 8 ICED 01 - C586/223
- 29. tested in these experiments. However, it was observed that the time to complete the task was significantly shorter (about 50%) in the clustered condition. Although this was not measured accurately, it may suggest that the list condition designers required more time to gain an overview of the space than the designers for whom the solution space was spatially laid out. The overwhelmingly popular solution represented three sub-mechanisms that give key two element combinations that re-occur. These included mechanisms that allowed orthogonal changes in force direction, such as lever to lever; lever to cam; or lever to screw links. The reports collected from these designers suggested that the criteria applied was simplicity of solutions, ignoring repetitions of sub-mechanisms and spatially translating force to force elements such as tie-rods. The effect of prompting the designers for more solutions in these cases led to the choice of variants on these mechanisms with tie-rods added to circumvent possible spatial constraints. The limitations of the evaluation lie in the small number of designers sampled and the use of a single problem. In addition the previous evaluation [2] had established a commonalityof 75% between the software-clustered groups and the designers intuitions as indicated by a paper- based solution-sorting task. In addition, the clustering method was based on a feature based on a count of pairs of solutionelements. Other features of the solution chain could be used for clustering to further test the effectiveness of the clustering approach. It was also clear from the designers performance that a more realistic task would involve a realistic design task carried out over a longer period. In addition to this, the pseudo 3D visualisation software was not integrated into the DESYN interface such that visualisation of individual solution chains was possible during the trial. It was evident that such a visualisation tool would have assisted the designers understanding of the individual solutions. Work is currently underway with a larger group of designers, sampling a wider range of problems over a range of sizes of mechanisms and difficulties of task. 6 Conclusions and further work A software system for assisting the visualisation of alternative solutions to mechanical design problems involving transmission and transformation of mechanical forces and motions was successfully implemented. This utilised a clustering algorithm for the purpose of data reduction and visualisation of a large solution space. An empirical evaluation of this system examined whether the designers' sampling of the known solution space was effectively assisted by an element-pair clustering represented in a 2D display of clusters, in conjunction with a pseudo-3D visualisationtechnique. There was no evidence for any improvement in the number of clusters sampled when the interface provided a list of solutions as opposed to a 2D cluster membership diagram, though a substantial improvement in time to complete the task was noted. Two strategies were observed in the designers tested. The predominant strategy involved listing 'elemental' sub- mechanisms that were capable of orthogonal changes in force direction. These designers ignored repetitions of sequences or complex combinations as well as elements providing linear or parallel force transformations. These second strategy group appeared to list solution sets that were clearly distinctive sequences, because of simplicity or unique element combinations. The small number of designers sampled does not enable generalisation to designers at large as the results may reflect individual strategies. It is also possible that designers' choices were ICED 01 - C586/223 9
- 30. affected by the lack of a genuine engineering task in the short trials. Firmer conclusions will be facilitated by the collection of further data using an enhanced assistive interface and more realistic and rigorous design tasks. 7 References [1] Chakrabarti, A. and Wolf, B., Reasoning with Shapes: Some Observations from a Case Study", Proc. 1995 ASME Design Engineering Technical Conferences: (9th Intl. Conf. on DTM), DE-Vol. 83, Vol.2, pp-315-322, 1995. [2] Fricke, G., Experimental investigation of individual processes in engineering design, Research in Design Thinking, (N.Cross, K. Doorst and N. Roozenburg eds.) Delft University Press, pp 105-109, 1992. [3] Chakrabarti A., and Bligh, T.P. "An Approach to Functional Synthesis of Design Concepts: Theory, Application, and Emerging Research Issues", AI in Engineering Design, Analysis and Manufacturing, Vol. 10, No. 4, pp-313-331, 1996. [4] Langdon, P., and Chakrabarti A. "Browsing a large solution space in breadth and depth", A. 12th International Conference on Engineering Design (ICED 99), Munich, Germany, v3 p. 1865-1868, 1999. [5] Johnson,B., & Shneiderman, B. "Tree-maps: A space-filling approach to the visualisation of hierarchical information structures". Proc. IEEE Visualization'91. IEEE, Piscataway, NJ, 284-291, 1991. [6] Kaufman, L. & Rousseeuw, P.J., Finding Groups in Data. Wiley Series in Probability and Mathematical Statistics. John Wiley and Sons, Inc., 1990. [7] Campbell, M.I. ,Cagan, J., Kotovsky, K., A-Design: An Agent-Based approach to conceptual design in a dynamic environment. Research in Engineering Design, Vol. 11, pp 172-192, 1999. [8] Finger, S., and Rinderle. J.R., A transformational approach to mechanical design usinga bond-graph grammer. In Design Theory and Methodology, DTM 89, vDE Vol. 17. pp 107-115, 1989. [9] Bracewell, R.H., Sharpe, J.E.E. Functional descriptions used in computer support for qualitative scheme generation - Schemebuilder. AI EDAM Journal - Special Issue: Representing Functionality in Design 10(4): p. 333-346, (1996). [10] Robertson G.G., Card S.K., Mackinlay, J.D., Information visualization using 3D interactive animation. Communications of the ACM, Apr 1993, Vol.36, No.4, pp.57-71, 1993. [11] Mackinlay,J.D., Rao.R, Card, S.K., An Organic user interface for searching citation links. Human Factors in Computing Systems (CHI) Conference Proceedings, Vol.1, pp.67-73, ACM, New York, NY, USA. 1995. Corresponding author's name: Dr. Patrick Langdon Engineering Design Centre, Cambridge University Engineering Department, Trumpington Street, Cambridge CB2 1PZ, UK, Tel: +44 1223 766961 Fax: 444 1223 332662, E-mail: pml24@eng.cam.ac.uk, URL:http://www-edc.eng.cam.ac.uk/people/pml24.html © Cambridge Engineering Design Centre2001 10 ICED 01 - C586/223
- 31. INTERNATIONAL CONFERENCE ON ENGINEERING DESIGN ICED 01 GLASGOW, AUGUST 21-23, 2001 SUPPORTING NON-STRUCTURED GRAPHICALINFORMATIONIN INTEGRATED DESIGNTEAM E Blanco and M Gardoni Keywords: design information, design understanding, hypermedia and multimedia, knowledge management The aim of this paper is to focus on the use of draft and sketches in the group understanding within Integrated-Team. We suggest specifications of a communication tool that supports and structures numerical drafting and treatment for capitalisation of those drafts connected to messages. 1 Evolution of Engineering Information Requirements Traditionally, engineering activities are performed in a sequential order. Over the last ten years, companies have tended to apply the Concurrent Engineering approach (CE) in order to drastically reduce the time-to-market of their products [1]. This approach is opposed to the sequential engineering approach because they have respectively two fundamental ways of working. In sequential engineering approach, the work starts at the reception of the results of the early stage when CE is based on information exchanges. Those exchanges are mainly performed verbally face-to-face or on the phone. Unfortunately, this information is therefore poorly controlled. This term "control" can be characterised in terms of four criteria [2] : - Information Structuring: it relies on a formal conceptual scheme which organise information at appropriate places. The evaluation criteria are the easyness to organise information in an intuitive and logical way and to be able to manage information at separate locations. Information Sharing: ability of "pushing" information. - Information Access: ability to "pull"information. - Information Capitalisation: ability to store and process information for later re-use, we assume that it must comply with the cycle of 'company knowledge capitalisation', i.e. locate, memorise, use information and update information. Taking into account these new requirements, we characterise the type of information exchanged in Integrated Team. We will focus here on information Structuration model. In order to meet the rigour needs of companies without going down on a too fine level of granularity, we choose an instructional design of the information significance. We then consider that the construction of a sentence corresponds to combine instructions formulated in term of variables, which provide a sense to the statement. Exchanged information is then an abstracted entity, a theoretical object which consists of [3]: ICED 01 - C586/418 11
- 32. - Linguistic components which build the significance of information starting from instructions. - Rhetoric components which bring a sense to information by addition of contextual information. Figure 1. Instructional design of the information significance Thanks to this instructional design of the information significance, we characterise different types of information [2] : Structured-Information (SI), linguistic and rhetoric components of the 57 are generally imposed. - Semi-Structured-Information (SSI), linguistic components of the SSI are little formalised and rhetoric components could be parsimonious. Non-Structured-Information (NSI), the NSI are very little formalised and the rhetoric components can be very light if they ensure a sufficient degree of relevance for the comprehension of information by the receiver. The MICA approach [4] (a specific interactive messaging system) has been developed and experimented in order to harness of NSI and to capitalise on relevant information exchanges. Also a Groupware tool called MICA, was created through an Intranet. It has been implemented and put into operation in an engineering team of twenty people in May 1998. Some return on experience has already shown improvement of the efficiency of the Integrated Team. Nevertheless, this kind of capitalisation from linguistic data is limited because of the graphical NSI lack. Sketching takes a large place in the group understanding of technical problems. Goel [5] also argues that freehand sketches facilitate creative and explorative work at the early stage of design, because they are ambiguous semantically and syntactically dense. Sketches could be considered as Graphical Non Structured Information (GNSI). They are at the same time models of the product and communication vectors. MICA only takes into account linguistic data but does not take care of GNSI management. This is today one of the actual limit of the MICA approach and more generally of Knowledge Management systems [6]. Generally Groupware solutions for interactive sketching have to be improved. 2 GNSI in collaborative design In order to analyse the role of GNSI in collaborative design we have carried out a design experiment [7]. This experiment was based upon the distributed design model [8]. Three roles 12 ICED 01 - C586/418
- 33. were represented, the functional, structural and machining roles. A camera has recorded the whole progress of the design. The work began with the requirement list supplied by the client a few days before the experiment. It took six hours to implement the detailed drafts of each part of the product in order to manufacture it. 2.1 Sketches as IntermediaryObjects of the designprocess In order to observe and analyse the design processes, we suggest to start with the Intermediary Objects (IO) which, circulating at any given moment between the actors of the process, can be seen as resulting from their design work but also as supporting and highlighting it [9], [10]. In modelling the future product they also act as communication vectors between the product designers. These two aspects are so indissociable in the reality of the process that we cannot isolate one from the other without deforming their nature. Because of this hybrid nature, intermediary objects are "analysers", making it possible to describe the actual design process. All along the design task, the object can not exist without the actor and the actor without the object. What we observe, see and feel in the design process, are objects being constructed, talked about, manipulated, interpreted, transformed, etc. Considered separately, the objects and actors are merely capacities for action. They are inert, static, mere?? "possibilities". It is the action, or rather the inter-action in which they are engaged that endows them with force, meaning and effective reality. The intermediary objects are also mediating objects: first of all, because of the mode of representation they use. This leads to a particular way of objectivising the idea or the intention by inscribing it in a specific organised matter. Thus, for example, a drawing is not simply a faithful representation of the mental idea or specifications : it is a translation, i.e. a realisation and a transformation which has been carried out according to its specific constraints (including those of the material used, which may vary depending on whether it is on paper or on screen) and its own rules and conventions. Lastly, the way in which the Intermediary Objects play the role of mediators in the design process is a result of their being representations. IO contents are mainly of cognitive nature : it results from the use of acquired knowledge by the actors in their various fields. They also signal the gradual production of new knowledge about the product as it is being designed. However, these contents are only of a cognitive nature in a situation of action, oriented by a given project and made up by the choices of the actors, the multiple negotiations that they are involved in, and the compromises and decisions that result from these. Therefore the cognitive contents cannot be isolated from the context of action 2.2 Classification of sketches in the collaborative process The role of the objects like sketches in collaborative design process are quite complex. Ferguson [11] identifies three kinds of sketches : thinking, talking, prescriptive. The situation of the experiment that we have carried out emphases on the second category of sketches. The collaborative activity involved many talking sketches. On the contrary, studies based on single designer activity mainly point out the thinking role of sketches [12]. In the experiment, we had characterised sketches [6] through an axis from open object that are able to support negotiation, to closed objects that mainly support prescriptions. We observed ICED 01 - C586/418 13
- 34. that the same object can support different status even if some objects get materials and cognitive characteristics that enhance opened or closed use. The situation of action as well as the material and cognitive properties of the IO influence the status. In the protocol, for example, we observed that the same sketch can be realised by an actor for his own use and after a few minutes be proposed to the others for evaluation. It moves from a private status (it is a thinking sketch), to a public one in the centre of the table. The existence of the private area is very important for the designer. This area is the thinking area that allows to explore ideas without judgement from the other. In a second step, the actor presents the meaning of his sketch and the other actors can assess the solution in their own point of view. This object then acts as a conjecture of solution. We can say that it is an externalisation of the designer's solution mental image. But this talking phase is also a thinking phase for the group that can make the sketch evolve. Every sketches are not built in a private area. Some are drawn directly in the public area of the table. A Sketch can also move from a talking to a prescriptive status. A sketch which has been the instrument of solution negotiation and elaboration, can get a prescriptive status after the group has built an agreement over it. For example, the sketch figure 2 got this prescriptive status by the group agreement. This decision is confirmed by a mark on the object: One of the actor underlined the word piston. This mark (detail Al) acts as the validation of the agreement. The definition and the negotiation of this part is then closed. Goel [5] suggests another way to characterise the evolution of sketches in he design process: the transformations typology. He identifies two types of operations occurring between successive sketches in the early stages of design: lateral and vertical transformations. A lateral transformation movement goes from one idea to a slightly different idea. Vertical express a movement from one idea to a detail or an extract of it. Then we can say that the role of the sketches depends on their status and the evolution between to successive sketches. The next table sum up the different categories of sketches we have identified. Thinking Talking / open Prescription / closed Lateral transformation Private new conjecture Public new conjecture New prescription Vertical transformation Private in depth conjecture Public in depth conjecture In depth prescription The status of the sketch is important in the sense that they partially characterise the situation of action where the actors are involved. Referring to the model presented figure 1, we can notice that the perception of the situation allows the receiver to build the sense of the information within the situation S. It act as a part of rhetoric components. 14 ICED 01 - C586/418 Table 1. Typology of sketches Private area Public area
- 35. 2.3 Sketches as a process of building conventional support For the explanation of this point, we must focus on the building process of those objects. The Object presented in figure 3 is in fact an addition of different sketch steps. The building of this sketch represents 20 minutes of group interaction during the design process. We consider that it is in this object that emerges the solution principle developed during the next phases of the experiment[7]. The first sketch (detail A) was drawn to put a end to a misunderstanding between two actors. They did not have the same interpretation of the requirements. The three sketches (A, B, C) allowed them to build a shared understanding of the problem. Then, from this representation a third actor had doubt about the connection between the device and its environment. This actor assessed the rubber tubing connection (detail D) as a non reliable solution for this system. From this negative evaluation of the solution elements represented on this sketch emerged a few minutes later a rigid and quick connection solution that is represented (detail E) on the sketch. This move is a lateral transformation that highlight the emergence process. Then this solution offered new opportunities for the inter-connections system of blocks. Finally a general overview of the system was drawn to visualise the ability of blocks interconnection (detail F). This last view is a vertical transformation showing the same solution from another point of view. Figure 2. Status of object moves during the process. Figure 3. Different level of sketches in the same draft... This sketch shows that the knowledge of construction process is necessary to be able to understand the sketch. In this sketch, aborted solutions and the solution chosen are represented on the same level. Nothing but the knowing of the process allowed the actors to differentiate the good from the aborted solution. This point allows us to point out that this type of GNSI is not able to support a prescription task. It failed as a memory of the process, In our protocol, we highlight that the actors themselves had difficulties to re-use their own sketches out of the action process itself.[7] The actors created, during the process, some symbolic devices which were connected to semantic in a local convention. This allowed them to let some part of the device in a low definition level: fuzzy and partial. The objects were used as "cognitive artefacts". The participants did not describe precisely what they were talking about, they showed the different elements they wanted to highlight with a finger or a pen and they used diectic words to point them. For example, an actor said: "that will be difficult to manufacture" (showing the elements concerned with his pen) ICED 01 - C586/418 15
- 36. We want to highlight the fact that the sketches act as pragmatic conventional supports [13]. Those conventional supports are negotiated in the interactions between actors, even if they can also use a higher level of convention as cultural ones, shared by the actors (the rules of industrial drawing in our case). The sense of the sketch is built in the action process and the conventions which allow this interpretation are quite local. Rhetoric components of sketches are specific and part of the linguistic components are built within the action itself. This characteristic of GNSI is important in order to imagine design tool able to support GNSI in an asynchronous communication process; furthermore, in order to involve sketches in a knowledge capitalisation process as MICA does for textual NSI. Indeed local conventions which allow interpretations are built in the interaction process. They won't be available to actors that were not involved in the process itself. 3 Towards a communication tool MICA GRAPH The first quality of a sketching device for designer seems to allow fast freehand sketch. Previous studies have highlighted that the cognitive tasks of manipulating the sketching tool have to be minimised [14]. Even if we can expect that the actors will learn the tool. That is why the tool is based on a simple blackboard technic and uses a digital table with an electronic pen. Indeed, when the designer needs formal drawing he will mainly use CAD or Structured information tools. Sketches are just GNSI and support fuzzy and partial definitions of elements. The functions that we offer are not supposed to help the drawing process but to partially structure GNSI in order to capitalise and treat them in the knowledge process. We are at the moment unable to manage a knowledge treatment directly on sketches. The principle we develop here is to characterise GNSI, track the steps of the building process and to associate textual and symbolic information that the data-mining techniques used by MICA are able to treat. At the moment, a designer creates a new sketch, he creates a properties file containing the legend of the file, type of the GNSI created and all the textual annotations contained. All this textual information is able to be treated by the MICA data-mining process [4]. 3.1 Tracking the different steps of the object We have highlighted the difficulty to track the different steps of a sketch out of the drawing process. It is also impossible to identify in a sketch the valid elements from the aborted solutions. It is important to follow the sketches modifications and to track the different sketch levels and steps present in the same final draft. We also want to know who had drawn on it. We propose to develop a structure of layers in order track the process. Each actor who wants to draw on an existing draft has to open a new layer. He can chose two options: "transparency" that allows him to sketch on the previous element or "blank" that opens a new layer disconnected from the previous sketch. In order to track information about the sketch contents and intentions, we characterise the type of opened layers. Six types of layers are available corresponding to the typology proposed in table 1. Public layers and private layers have different properties. When an actor wants to send a sketch, the public layers are automatically sent. Private layers sending is optional. If the actor accepts to send them the category changes moving from private to the public area. 16 ICED01-C586/418
- 37. In order to highlight decisions concerning a specific area of a sketch, actors can modify the colour of this area. By drawing a boundary line around this area, they notify that it is fixed. This materialises the compromise within the object. 3.2 The necessity to associate annotations on GNSI In order to allow the actors to interpret the sketches in an asynchronous communication mode, and to capitalise textual information about the content of the sketch, we propose different devices to the users. Firstly an actor as to define a title to the sketch using few keywords. We notice here that some contextual components (product, process, state, part etc. ) are already identified in the MICA form.[2] [4] It is important to give the actors the ability to express comments in the sketch itself. In a synchronous interaction around sketches actors explain some elements of the solution they focus on. Drawing is generally accompanied by a speech concerning the sketch. In an asynchronous process, the speech does not exist. It is supported by writing in the message or by a local annotation on the sketch. In the same way, actors need to be able to focus the attention of the receiver on a specific area and to associate elements that allow the receiver to interpret the sketch in the right way. They have sometimes to explain some symbolic elements they have used to fuzzy represent an element. Annotations can allow them to do it. The annotations are textual information that are connected to the graphic properties. A further step is to allow an actor to define different types of annotations using colour and type of markers. Each time he defines a new type, he has to explain the legend of this symbolic feature. Laureillard had pointed out the importance of what he called co-operative feature in the co-operation between specialists [15][16]. Those symbols are local conventional supports which refer to shared knowledge between the different actors. We don't have deeply investigated the different types of annotation. Our proposition is to offer a toolbox to the actors. They have to define themselves the relevant type co-operation feature. 4 Conclusions The concepts developed in the MICA-GRAPH tool have to be improved in design practices. The aim is not to suppress synchronous interactions but to complete the offer of asynchronous communications tools for design activities. Engineering design needs visual representations: structured (like CAD models) and non structured (like sketches). GNSI allows to manage quick conjecture-evaluation process. We notice that the move from direct interaction to textual explanation implies the risk of the deterioration of the contents. But the advantages of textual information related to GNSI aretheir ability to be treated in a knowledge process. References [1] B. Prasad, Concurrent Engineering Fundamentals - Integrated product and process organisation, Vol. 1, Prentice Hall, Englewood Cliffs, NJ, 1996 [2] M. Gardoni, M. Spadoni, F. Vernadat, "Information and Knowledge Support in Concurrent Engineering Environments", 3rd International Conference on Engineering Design and Automation, EDA'99, Vancouver, B.C., Canada, August 1-4, 1999 ICED 01 - C586/418 17
- 38. [3] Moeschler, J. Modelisation du dialogue (representation de 1'inference argumentative), Editions Hermes, Paris, 1989 [4] M. Gardoni, Maitrise de 1'information non structuree et capitalisation du savoir et du savoir-faire en Ingenierie Integree - Cas d'etude Aerospatiale Matra, PhD thesis of Metz University 1999 [5] Goel V, Sketches of thought MIT press Cambridge, MA 1995 [6] Gardoni M., Blanco E Taxonomy of information and capitalisation in a Concurrent Engineering context 7th ispe international conference on concurrent engineering (CE'2000), Lyon, France, July 2000 [7] Blanco E., 1'emergence du produit dans la conception distribute PHD thesis INP Grenoble 1998 [8] Garro O., Salau I., Martin P., Distributed design theory and methodology, Concurrent engineering: research and applications, vol 3/1/1995. [9] Vinck D., Jeantet A., 1995 Mediating and commissioning objects in the sociotechnical process of product design: a conceptual approach, pp111-129 in D. Mac Lean, P. Saviotti, D. Vinck (eds), Management and new technology : Design, Networks and Strategy. COST Social science serie. Bruxelles. Commission of european [10] Blanco E., Garro O., Brissaud D., Jeantet A., Intermediary object in the context of distributed design CESA Computational Engineering in systems applications, IEEE- SMC, Lille, July 96 [11] Fergusson E,S, Engineering and the Mind's Eye MIT press Cambridge, MA 1992 [12] Rodgers PA, Green G., McGownA. Using concept sketches to track design progress in Design studies 21 N°5 pp 465-481 sept 2000 [13] N. Dodier, les appuis conventionnels de I'action elements de pragmatique sociologique Reseaux n°62 edition CNET, 1993 [14] Aytes G., Comparing Drawing Tools and Whiteboards: An Analysis of the Group Process in CSCW 4: 51-71,1996 [15] Laureillard P., conception integree dans I 'usage, PHD thesis INP Grenoble 1999 [16] Boujut, Blanco The role of objects in design co-operation: communication throught virtual or physical objects. In COOP 2000 conferences Workshop Sophia Antipolis May 2000 Dr Blanco Eric Soils Solids Structures laboratory, Domaine Universitaire, BP 53, 38041 GRENOBLE Cedex 9, France, Tel: (33)-4-76-82-70-11, Fax (33)-4-76-82-70-43, mail - eric.blanco@hmg.inpg.fr Dr Gardoni Mickael GILCO laboratory, ENSGI, 46 avenue Felix Viallet, 38031 Grenoble Cedexl, France, Tel (33) (0)4 76 57 43 33, Fax (33) (0)4 76 57 46 95, mail: gardoni@gilco.inpg.fr (c)IMechE2001 18 ICED 01 - C586/418
- 39. INTERNATIONAL CONFERENCE ON ENGINEERING DESIGN ICED 01 GLASGOW, AUGUST 21-23,2001 AUTOMATIC COMPOSITIONOF XML DOCUMENTSTOEXPRESS DESIGN INFORMATION NEEDS A Dong, S Song, J-L Wu, and A MAgogino Keywords: Information analysis, classification and retrieval; information representation; design information management 1 Introduction Engineering design is an information intensive activity. It is reported that designers spend in excess of 50% of their time in handling information [7]. Thus, the efficiency and the quality of the design process depend considerably on how well designers are able to handle large amounts of information. One study [8] of the information required by design engineers to complete their jobs indicated that less than 50% of that information was actually available and only 20% could be provided by the existing specialized applications. Directing the right information to the right person at the right time is a complicated but crucial task. Design information management has received increasing attention in recent years as a result of these findings and the recognition that lacking sufficient or missing key design information may lead to sub-optimal decision-making and design [2,4]. Much of the existing research has focused on the capture, storage, indexing and presentation of design information including informal information [3,9]. Less work has been done on information retrieval based on an understanding of individual designers, their experience, their skills and the ways in which they use information in the context of their design task [1]. One key step in finding the right information is expressing information needs in context. This paper presents a methodology to generate an XML (extensible Markup Language) document that expresses the information needs of a design engineer. XML documents and their underlying Document Type Definition (DTD) offer an efficient structure for the organization of design information [5] and representation of information needs. Through declarations of XML entities and the inherent structural hierarchy of XML documents, XML documents can express the designer's information needs while framing the design's structural hierarchies. For example, an element in the DTD may permit the design engineer to express a preference for formal company documents of past designs (e.g., technical memos) rather than informal design notes. The data in the XML document may be drawn from a repository of semi-structured or unstructured text documents that the design engineer has retrieved and placed into a personal information store by using an information retrieval system. Our methodology draws from the computational linguistics techniques of natural language processing and latent semantic analysis (LSA). We assume there exists some underlying information needs that are expressed by the type of information the engineering designer wishes to view and download into a personal information store. The "type of information" will be distinguished primarily by subject but may include other identifiers such as the format of the information and the intended audience. By applying these linguistic techniques, we can ICED 01 - C586/422 19
- 40. construct an XML document that is descriptive of this underlying need and contains information directly from the documents that is consistent with the major patterns of information preferences of the designer. 2 Methodology 2.1 Technical Approach The methodology for automated composition of the XML documents proceeds along two axes: 1) explicitly solicit information needs from the designer through standard information retrieval means; 2) implicitly monitor the information retrieval behaviour of the designer to information sources including the type, quality and information contained in the documents the designer chose to retrieve. We test this methodology on access to unstructured engineering data, such as full-text, because unstructured, textual documents are the principal mode of communication by engineers. Before proceeding to a detailed discussion of our methodology, we discuss two core technologies to the methodology: mining of transaction logs to learn information needs implicitly and computational linguistics. 2.2 Learning Information Needs Implicitly Many difficulties exist in determining what information a person wants to see as well as modelling user information needs. Most information retrieval systems require that people express information needs through a set of keywords or key phrases. However, ascertaining information needs simply by a word or two is inadequate and subject to loss of contextual information. For engineering design, studies have shown varying information needs of designers depending on level of expertise and stage of the design [1,6]. Our approach in modelling user information needs is based on a human-centred computing. Our methodology examines the user's document access patterns, that is, the user's personal information store and transaction history in an information retrieval system, for patterns of information preference. The basic approach is to examine the user's session information over all sessions while using the information management system. In learning information needs implicitly, we are primarily interested in discovering the similarity between documents that the user has downloaded into a personal information store. The assumption is that the user's particular choices of documents to store locally are indicative of information needs. Thus, instead of requiring the designer to a priori categorize the information, the system attempts to learn a similarity mapping using contextual clues such as project name, engineering discipline, and document format. Similar categorisation strategies have also been found in the classification of supplier information practised in industry [12]. To ascertain relevance, the system records each document downloaded into the user's personal information store. This is an accurate indicator of relevance because, in our system, before the user may download the document, the user has already read a brief description of the content of the document containing meta-information such as abstract, author, document type, and subject. The assumption is that during any single session utilizing the information retrieval system, the user has a dominant goal (information need) in mind that is expressed by the type(s) of documents the user decides to download. Similar assumptions exist in other studies [13] of information retrieval systems. We use the vector space approach [14] for document and query representation. We analytically modelled information needs as a linear combination of the vectors representing 20 ICED 01 - C586/422
- 41. the query and the relevant document. The weighted vector average (arithmetic mean) of all combined vectors consisting of the query and relevant documents is called the "centroid" of the user's intended information needs. The centroid is then used as a representation of the dominant information need of the user. The maximum angle between the document vectors or query string vector and the centroid is used as an indication of the variation in the user's information needs. 2.3 ComputationalLinguisticApproaches Our methodology employs two computational linguistics techniques, natural language processing and latent semantic analysis, to extract and summarizethe primary topic of a set of similar documents. Natural Language Processing While we use the designer's past transactions as an indication of information need, the structured information contained in the documents that are referenced in the transactions needs to be discovered and refined before it can be utilised effectively. The key phrase retrieval process helps in crosschecking the indexed subjects and compacting the size of the aggregated XML document by representing paragraphs of text with just a few representative noun phrases. As the designer adds documents to a personal information store, we do not need to concatenate all the textual information about the document to the XML document, just the extracted noun phrases that are not already in there and plug them under the appropriate tags. By identifying the noun phrases in the document, we are able to find the corresponding contents for each XML tag in the text. Additional rules and procedures are needed other than standard noun phrase retrieval process to perform this task for all tags. This latter component forms a part of future research. Extracting from the full-text content-bearing noun phrases documents that can be used in profiling and indexing involves 3 steps: tokenization, part-of-speech (POS) tagging and noun- phrase identification. Tokenization is a procedure that identifies sentence boundaries and removes extraneous punctuations. POS taggers then take the processed corpus and tag each word with their POS information. Taggers that operate following semantic rules or just statistical information were developed. After the text corpus has been tagged with POS information, we could use the contextual information to identify noun phrases. The extracted noun phrases are then attached to the corresponding DTD elements of the document. Latent Semantic Analysis Latent Semantic Analysis (LSA) [9] is a statistical model of word usage that permits comparisons of semantic similarity between pieces of textual information. The idea is that the totality of information about all the word contexts in which a word does and does not appear provides a set of mutual constraints that largely determines the similarity of meaning of words and sets of words to each other. The primary assumption of LSA is that there exists an underlying or "latent" structure in the pattern of word usage across documents. LSA uses the matrix technique of singular value decomposition (SVD) to reflect the major associative patterns of words in the document and to ignore the smaller influences. The ability for LSA to remove the obscuring "noise" makes LSA useful as an analytical tool for discovering the primary conceptual content of documents. We use LSA to help us to categorize and group the documents by topic material that the user has downloaded and revealed as relevant and useful. Once these principal groupings are identified, we can then apply the rest of our methodology ICED 01 - C586/422 21
- 42. to express information needs for each "centroid" of documents within a semantic locality through a single XML document. 2.4 ComposingXML Documents The system initiates by asking the user to specify an XML DTD containing metadata elements (XML entities) that express information needs. For this study, we asked that users express their information needs using a well-known metadata set, the IEEE Learning Object Metadata (LOM) [10], and in particular the core 20 elements. The LOM defines the information required to manage, locate and evaluate learning resources. Having a standards-based XML DTD allows us to assess our methodology's completeness and efficacy and for comparison against other systems. Because of the analogous human cognitive processes of information processing and learning [11], the LOM serves as a reasonable model for describing information needs. Once the user has specified an XML DTD, the user must now utilise the information management system, retrieving and downloading design documents. The system implicitly monitors the information retrieval transactions of the user, eventually formulating a seed an XML document containing information from the user's session. Typically, the XML document is seeded with the initial query the user posed to the system. The implicit stage contains two phases. In phase one, the system applies latent semantic analysis to characterize the knowledge conveyed by the all documents the person chose to view. To perform this phase, we analysed the transaction logs containing the transactions of both the current user and all other users of the system. The user's query and document(s) downloaded are recorded for each visit. The latter information identifies the relevant documents necessary for phase two. Using a similarity measurement, the system identifies topics represented by the documents the user viewed. By doing this step, the system ascertains the topic locality of the various documents the person viewed. This is a critical step because the user may have multiple and widely varying information needs. Then, for each topic locality, the system augments the XML document expressing the user's information needs with the metadata elements from each relevant document. In practice, the information management system will contain most of the information required to complete the tagging such as Author, Title, Date, and Format. Subject and Description information are generated automatically from the second phase. This matching can be done by exact one-to-one correlation, i.e., both the tag and attribute match, or via a crosswalk between the information about the document contained in the information management system and the XML DTD. Both of these techniques will have required some prior means of tagging the documents in the information management system. In the second phase, we apply natural language processing techniques to ascertain the principal subject of the documents within each topic as discussed in Section 2.3. The process repeats until all possible elements in the user's original XML DTD are filled, resulting in a fully marked up XML document. The completed DTD is now an expression of the information needs of the designer, based upon the available information stores and the pattern of information retrieval undertaken. In summary, a designer's information needs can be found implicitly by looking at the set of documents the designer has deemed relevant and useful and stored in a personal information store. We use LSA to find signatures of similarity in this set of documentation. Once we find the signatures, we look at what the original information needs were to find the centroid of the 22 ICED01 - C586/422
- 43. similarity. Finally, we construct a compound XML document that essentially reconstructs the full LSA space by combining information from all the similar documents, with some of the information filtered through NLP techniques to reduce the size of the document. Other information about the document indexed in the document database, such as document type, is added to the corresponding XML tag. This final XML document is then an expression of the designer's information needs. 3 ExperimentationandResults 3.1 Test Case The experiment and prototype evaluation was conducted on a digital library project for science, math, engineering and technology education. Students and educators use the digital library to download courseware into their personal information stores. The documents used in the study discuss the design of engineering devices and related scientific theories. The users of the system typically search for material on engineering education. This is their primary information need. 3.2 Results First, we validated the ability of our methodology to discover information needs. We conducted this study by analysing for the known information needs of all users of the digital library. Based on the fairly homogeneous content of the digital library (courseware on engineering design) and the known profile of the audience of the digital library, we expected that our methodology would reveal one dominant information need, namely courseware related to engineering education. We would not expect for the system to reveal numerous distinct clusters of information needs. We ran the latent semantic analysis over the entire usage database. Figure 1 illustrates the distribution of all users' information needs over multiple sessions. Figure 1. Information Needs Represented in LSA Space ICED 01 - C586/422 23
- 44. Each dot represents in LSA space the combined vector of the user's query and a downloaded document for one session. By visual inspection, one can note one dominant informationneed. Specifically, the most commonly downloaded documents by all users of the system were case studies and courseware on the design of disk drives, a specific subset of engineering education. This result corroborates the known informationneeds of users of the database. Second, we analysed for the information needs of individual users. Figure 2 illustrates the information needs of one sample user, specifically information on "control systems". The circle represents, in LSA space, the initial query to the information retrieval system whereas the boxed numbers indicate the documents downloaded by the user. One can then apply latent semantic analysis to ascertain the similarity between downloaded documents, the original query, and the documents themselves. Users may have multiple information needs despite using the same keyword to query the information retrieval system. In the example shown, document 165 is relevant to the user's query but not similar to the other documents, therefore potentially indicating different informationneeds. For this document set, we found that an angle of 71o between documents provided an adequate measurement of similarity. Based on the set of similar documents, we computed the centroid. Figure 2. One User's Information Needs Finally, we generated the XML documents to express the information needs. Portions of an XML document are illustrated in Figure 3. <!-- The Core IMS Learning Object Metadata in XML, a subset of the IEEE LOM V3.5. --> <metametadata> <metadatascheme>IEEELOM:1.0</metadatascheme> <language>en-US</language> </metametadata> <general> <title> <langstring> educational software engineering graphics tutorials engineering visual encyclopedia 24 ICED 01 - C586/422
- 45. mechanics virtual disk drive design studio </langstring> </title> <language>en-US</language> <description> <langstring> acme disk drive company <lifecycle> <contribute> <role> <langstring lang="en">Author</langstring> </role> <centity> BEGIN:vCard <?xml> Figure 3. Sample XML Document The XML document expresses out in human-readable format a summary of the user's information needs as an aggregation of the documents that the user found relevant and useful and that were related to each other. 4 Conclusions This research has established a basic framework for identifying and modelling engineers' information needs using XML documents. We performed latent semantic analysis over a collection of engineering resources to construct information needs as vectors in LSA space based on usage analysis. We visualized different information needs in multi-dimensional space. Based on a cluster of similar documents representing an information need, we generated an XML document using natural language processing techniques to express the information need. These results are encouraging. They show that latent semantic analysis can be applied to the task of ascertaining information needs by monitoring the information retrieval habits of a user. In order to assess the actual "truth" of the XML document in representing the user's information needs, we would need to have the user respond positively or negatively to suggested relevant information provided autonomously by the information retrieval system. We have projects in progress to incorporate this feedback. In addition, we are working on methods to incorporate reading time into the model of information needs and to predict the expected reading time of a document based on prior reading time. We expect this methodology to impact the use of information in design in several ways. First, the XML documents can be used as information filters to direct critical pieces of information to the designer as others generate them. In addition, intelligent software agents might use the XML document as a guide to search document repositories for new, useful information. The methodology may provide insight into the cognitive states of the designer over various stages of design, offering a tool to study how changes in information needs relate to the designer's understanding of the design problem. We are currently analysing the effect of time on information needs, particularly the rate of change of information needs. In addition, we are investigating learning the information needs of design teams by analysing team communication. Our methodology presents a new means for learning information needs through a combination of LSA, natural language processing and a human-centred approach which places emphasis on understanding what it is that the user is doing. ICED 01 - C586/422 25
- 46. Exploring the Variety of Random Documents with Different Content
- 47. Erdpech liegen wie Teppiche auf dem Ufer ausgebreitet, werden zusammengerollt und auf Prahme geworfen. Pechgestank erfüllt die Luft. Frauen in dunkeln Mänteln balancieren mit Pech gedichtete Töpfe auf dem Kopf, füllen sie am Kai mit Wasser und plaudern mit ihren Nachbarinnen, die eifrig ihre Wäsche in den Fluten des Euphrat spülen. Gleich unterhalb des Minaretts landete ich an einem offenen Uferplatz. Eine Schar spielender Buben stürmte bald herbei, und ein kleiner Türke in buntem Hemd rief mir zu: „Gestern ist Kut-el-Amara gefallen. Heute ist das Telegramm gekommen!“ „Bist du dessen auch sicher?“ fragte ich. „Ja“, antwortete er, „der Herr kann ja auf dem Telegraphenamt nachfragen.“ Schön, dachte ich, dann ist Bagdad außer Gefahr, auch der Weg nach Babylon noch offen, und begab mich mit meiner gewöhnlichen Begleitung, dem Gendarm und Sale, in die Stadt hinauf. Eng die teilweise mit Asphalt belegten Gassen, grau die Mauern, ärmlich die Lehm- und Steinhäuser. Dunkle Gänge führten durch offene Tore zu Hütten und auf schmutzige Höfe. Wasserträger mit tropfenden Ledersäcken auf dem Rücken, Esel mit Fruchtlasten, kleine Läden mit Sonnendächern oder offener Auslage von Brot, Erbsen, Granaten und andern Eßwaren; auf den Holzschwellen an den Gassen Kinder mit Schmutznasen, das Gesicht mit Fliegenschwärmen bedeckt; an einer Mauer Aussätzige von abschreckendem Äußeren — welcher Gegensatz zu dem lieblichen Bild, das Hit dem Ankömmling zu Wasser vortäuscht! Auf die braunen Fluten des Stroms öffnete sich in den winkligen Gassen nur selten ein flüchtiger Ausblick.
- 48. Gendarm Saalman. Zum Marktplatz mußten wir wieder hinabsteigen bis dicht an den Strand. Dort standen zwei stattlichere Häuser; in deren einem wohnte der Mudir von Hit, ein Araber, dessen fortschrittlich europäische Kleidung mit seiner bedenklich zurückgebliebenen Intelligenz auffallend kontrastierte. Das andere war das Telegraphenamt, wo mir der Fall von Kut-el-Amara bestätigt wurde. Der englische Oberbefehlshaber General Townshend war mit 13000 Mann gefangen — ein bedeutender Sieg also, dessen Kunde die ganze mohammedanische Welt durchlaufen und die Macht des Sultans kräftigen mußte. Am 29. April hatte der Feind seine Stellungen räumen müssen — zehn Tage vorher war Feldmarschall von der Goltz gestorben! Ein grausames Schicksal hatte es ihm verwehrt, diesen Siegestag zu erleben, den sein Genie und seine Umsicht an der Spitze der 6. Armee erzwungen hatten.
- 49. Landungsplatz in Hit. Während ich auf der Post weilte, war ein kleiner Doppelschahtur bei meiner Fähre angekommen mit einem katholischen Priester an Bord und einem zweiten jungen Deutschen namens Kettner. Wir verbrachten den Abend zusammen und verabredeten uns für den folgenden Tag zu gemeinsamer Fahrt. Hit verließ ich am 1. Mai, und nun näherte sich meine Stromfahrt ihrem Ende. Mein Freipassagier Asis war in Hit zurückgeblieben; daß er sich auf Französisch gedrückt hatte, wunderte mich nicht, wohl aber, daß er verduftet war, ohne ein — Trinkgeld für seine Reisebegleitung zu fordern. Das linke Euphratufer heißt von hier ab bei den Arabern El- Dschesire (Insel), ein Begriff, der sich ungefähr mit Mesopotamien deckt; das rechte Esch-Scham. Diese beiden Namen traten nun immerfort in Verbindung mit Strömung, Wind und Landungsplatz auf. Am Schamufer fuhren wir an den niedrigen Felsabhängen von Leguba entlang und lenkten dann hinter der tamariskenbewachsenen Insel Abu Tiban nach Dschesire hinüber.
- 50. Beim Palmenhain Tell-essued zwang uns heftiger Südost wieder längere Zeit still zu liegen, und voll Neid sah ich die stolzen Meheile in dem ihnen günstigen Wind mit vollen Segeln an uns vorüber stromaufwärts fahren. Während ein Araber gewandt am Stamm weiblicher Palmen hinaufkletterte, ihr Samengehäuse mit dem Staub der männlichen Blüten bestreute und dann die Blattbüschel mit Bast zusammenband, vertrieben sich meine Leute die Zeit mit Vogelfang und mit dem Bau kleiner Schiffchen aus Palmblättern. Dann versuchten wir die Fähre am Ufer entlang vorwärtszuziehen; bei offenem Strand ging es; wo aber Gestrüpp den Leinpfad unwegsam machte, kamen wir verzweifelt langsam vorwärts. Wir landen bei Tell-Essued. In dieser Gegend vollzieht sich der Übergang der Hochebene in das vollständig ebene Schwemmland, das zur Zeit der assyrischen und babylonischen Königreiche von gewaltigen Kanälen durchschnitten und mit üppigen Gärten und Äckern bedeckt war. Die weißen Kalksteinwände verschwinden, nur noch vereinzelte Ausläufer ziehen sich bis an den Strom heran. Der Horizont rückt in weite Ferne, und der Euphrat benutzt seine neu gewonnene Freiheit, um sich mächtig auszudehnen.
- 51. Unser Lager bei Tell-Essued. Auch am 2. Mai kämpften wir vergebens mit Gegenwind, und die zahlreichen Schöpfwerke — hier wieder von dem primitiven Typ — machten es fast unmöglich, die Fähre vom Lande aus weiter zu bugsieren. So krochen wir mit unendlicher Mühe bis Ramadije. An diesem Tag begegnete uns die erste Guffa, eines der runden asphaltbekleideten Korbboote von großer Leichtigkeit und Tragfähigkeit, die schon aus assyrischer Zeit bekannt sind. Der Ruderer war selbst Mast und Segel, er stand in der Mitte des Bootes, und sein Mantel war als Windfang ausgebreitet.
- 52. Meheile auf dem Euphrat. Am nächsten Tag fanden wir die Ufer auf weite Strecken überschwemmt. Der Euphrat macht hier große, oft fast kreisförmige Windungen, und das in diesen Schleifen liegende Land war von dem eigentlichen Strom kaum noch zu unterscheiden, so daß es die ganze Kunst meines Kapitäns erforderte, sich zwischen diesen unter Wasser liegenden Landzungen, Inseln und Schlammbänken, in kleinen und großen Seitenarmen zurechtzufinden. Da standen Palmenhaine mitten im Strom; dann wieder waren Äcker mit reifenden Ernten von der Flut verschont, und Schaf- und Rinderherden, Hütten und Zelte standen wie auf der Wasserfläche. Die Anwohner hier mußten geradezu ein amphibienartiges Dasein führen, jeden Augenblick gewärtig, von den launischen Wellen überrascht zu werden. Oft war auch das, was ich für Zelte hielt,
- 53. nichts weiter als die Segel der Meheileboote, die rechts oder links über der Wasserfläche emporragten und denen wir dann in großem Bogen, den der Euphrat beschrieb, begegneten. Meine Leute mußten schließlich nackt ins Wasser hinein, um uns nur vorwärts zu bringen. Dabei war die Insektenplage fast unerträglich. Wenn wir abends an einem öden Strand vertäuten, durfte ich mein Licht nur für mein schnelles Abendbrot brennen lassen. Überall schwirrte und surrte es von Insektenschwärmen, die in dieser Sumpfgegend und bei der tropischen Hitze myriadenweise gediehen. Das innere Dach meiner Hütte war mit einer schwarzen, kribbelnden Decke bezogen, plumpe Käfer stießen gegen die Wände, törichte Nachtfalter taumelten in die Flamme und plumpsten mit verbrannten Flügeln in mein Eßgeschirr. Die Grillen zirpten um die Wette, die Frösche quakten im Sumpf und das höhnische Lachen und langgezogene Heulen der Schakale ging in unheimlichen Wellen über die Steppe. Die schlimmsten Quälgeister aber waren Mücken und Moskitos, von denen immer einige durch die Maschen des Moskitonetzes schlüpften. Ein brennendes Jucken lief über den ganzen Körper, und an Schlaf war in diesem Bett von Brennesseln nur wenig zu denken. Am Morgen, wenn sie dickgefressen und zu faul waren, wieder hinauszufliegen, hatte ich wenigstens die Freude, blutige Rache nehmen zu können. Die Morgenkühle pflegte das Jucken zu beseitigen.
- 54. Dorf in der Gegend von el-Beschiri. Am 4. Mai erschien endlich über der glatten Steppe das Minarett von Feludscha, erst im Südosten, dann im Nordosten, denn der Euphrat macht hier wieder einen mächtigen Bogen nach Süden. Als wir den Landungsplatz erreichten, war das erste, was ich sah, ein Schahtur von Schrenks Batterie; die Abteilung des Roten Kreuzes war noch hier und sollte am nächsten Tag aufbrechen. Bei Feludscha ist der Euphrat ungewöhnlich schmal. Diesem Umstand hat der Ort seine Entstehung zu verdanken. Denn hier geht die große Karawanenstraße über den Strom. Die Brücke war aber des Hochwassers wegen eingezogen; die Pontons lagen am Ufer und sollten auch nicht mehr verwendet werden, da die Brücke für die Fähren nach Risvanije ein gefährliches Hindernis war. Die Reisenden müssen sich daher damit abfinden, daß sie und ihr Gepäck auf einzelnen Prahmen über den Euphrat gesetzt werden. Von Feludscha aus konnte ich nun zu Wagen auf der Karawanenstraße nach der Stadt der Kalifen gelangen oder zu Schiff bis Risvanije weiterfahren, das durch eine kleine Feldbahn mit Bagdad verbunden ist. Bei ruhigem Wetter rechnet man zu Wasser
- 55. bis Risvanije acht Stunden. Ich entschloß mich daher, auf der Fähre zu bleiben, mußte aber bald diesen Entschluß bereuen. Gegenwind und Gewitterregen, dazu ein tüchtiger Weststurm, zwangen uns immer wieder stillzuliegen und dehnten die acht Stunden zu mehr als einem Tag. Die Araber sind gegen nichts empfindlicher als gegen Regen; vor jedem kleinen Schauer ließ meine Besatzung die Ruder im Stich, flüchtete unter Deck und war erst wieder aufzutreiben, wenn wir zu kentern drohten. Einmal wären die Leute, als sie die Fähre ins Schlepptau nehmen wollten, beinahe alle im Schlamm ertrunken. Ich atmete daher erleichtert auf, als ich endlich am 5. Mai Risvanije erreichte und mein tüchtiger Doppelschahtur seine Reise von 1040 Kilometern ohne schwere Unfälle vollbracht hatte.
- 56. E Phot.: Schölvinck. Eine Guffa auf dem Tigris. Neuntes Kapitel. Mein Einzug in Bagdad. s regnete in Strömen, als ich meine Fähre verließ. Am Ufer erwarteten mich Herr Kettner und der Flugzeugmonteur Knitter, der zur Fliegerabteilung des Hauptmanns Niemayer gehörte. Ich kam gerade recht, ein Zug von zweiundzwanzig Wagen sollte nach
- 57. Bagdad abgehen, sobald der Regen aufhörte. Ich eilte daher zum Kommandanten von Risvanije, dem Kurden Ahmed Mukhtar aus Suleimanije, einem gewandten, höflichen Mann, der in der französischen Missionsschule zu Bagdad die klangvolle Sprache der Boulevards gelernt hatte. Alles ging nach Wunsch: der Zug sollte warten, bis ich reisefertig sei. Ich speiste also mit den Deutschen zu Mittag und trank Tee bei Ahmed Mukhtar. Mittlerweile packte Sale meine Sachen. Die Fähre, auf der ich unvergeßliche Stunden verbracht hatte, stellte ich unter den Schutz des Kommandanten. Vielleicht konnte sie zur Fahrt nach Babylon noch gute Dienste leisten. Das Gleis der Feldbahn führt bis zum Strand hinunter. Die Wagen sind aus Eisen und haben dieselbe Form wie die schwedischen Erzwagen auf der Linie Luleå-Reichsgrenze, sind aber viel kleiner. Die Lokomotiven waren noch nicht fertig; das einzige Zugmittel war Menschenkraft; Araber wurden zum Schieben der Wagen ausgehoben. Bis Bagdad wurden die Leute sechsmal gewechselt, jedesmal fünfzig Mann. Sie erhielten drei Brote am Tag und zusammen ein Lamm. Ihr Sold war unbedeutend, und oft genug rissen sie wieder aus. Demnächst sollten aber die Lokomotiven fertig und damit die Leistungsfähigkeit der Bahn bedeutend gesteigert werden. Jeder Wagen faßte zwei Tonnen. Die Güter, die mit meinem Zug befördert wurden, waren Kriegsmaterial und Proviant. Mein Gepäck wurde auf den letzten Wagen geladen; oben drauf thronten ich selbst und mein Diener Sale. Um 3 Uhr hatte der Regen aufgehört. Die Luft war frisch und kühl. Alle Mann standen an ihren Wagen bereit, und auf ein gegebenes Zeichen begann der wunderliche Eisenbahnzug sich in Bewegung zu setzen. Schnell ging es nicht, und wir als letzte mußten wohl oder übel jeden Aufenthalt mitmachen, den einer der vorderen Wagen verursachte. Zu beiden Seiten der Bahn breiteten sich Weizen-, Korn- und Haferfelder aus, denen ein Kanal Wasser aus dem Euphrat zuführte. Die ganze Gegend heißt Risvanije oder Nasranije, wie man es auch ausspricht. Der angebaute Feldstreifen war aber nur schmal, und
- 58. bald fuhren wir wieder durch die Wüste. Zu beiden Seiten der Schienen hatten die Füße der schiebenden Araber in den graugelben Alluviallehm Rinnen getreten, die jetzt voll Wasser standen. In der Nähe des Ufers waren wir durch eine kaum ein paar Meter ansteigende Höhe gefahren. Ein zweiter Hohlweg dieser Art bei Jusfije ließ vermuten, daß diese Höhen nichts anderes waren, als Dämme zu beiden Seiten uralter Kanäle. Links vom Wege lag, von einigen grauen Ruinen umgeben, die Grabmoschee Brahim-el-Halil Imam, auf der anderen Seite eine kleine Anhöhe namens Tell- achijen. Hier erhielt der ganze Zug Befehl zu halten, und mein Wagen wurde auf einem Nebengleis an die Spitze geschoben. So hatte ich nun freie Bahn und freie Aussicht, und wir ließen den übrigen Zug bald weit hinter uns. Die Bahn läuft schnurgerade nach Bagdad. Die ganze Entfernung beträgt nur 45 Kilometer; sie ist die kürzeste zwischen Euphrat und Tigris an der Grenze zwischen Mesopotamien und dem Irak. Kurz nach 5 Uhr erreichten wir die erste Station Kal’at Risvanije, wo neue Araber als Schlepper eintraten und ein neuer Gendarm zu uns stieß. Dieser berichtete mit bedenklicher Miene, in der vorigen Nacht sei zwischen der ersten und zweiten Station ein Zug von Räubern angefallen worden. Durch Gewehrschüsse habe man zwar die Angreifer in die Flucht gejagt. Immerhin sei es gut, an der gefährlichen Gegend so schnell wie möglich vorbeizukommen, denn man könne nie wissen! Unter diesen Umständen wäre es zweifellos vorsichtiger gewesen, bei den andern Wagen zu bleiben, die eine türkische und deutsche Eskorte hatten. Aber da ich den Arabern schon ein tüchtiges Trinkgeld versprochen hatte, wenn sie ordentlich liefen, mochte es nun auch dabei bleiben. Der Gendarm hatte es so eilig, daß er selber mit schob. Vor jeder Anhöhe aber rannte er mit einer unermüdlichen Ausdauer, obgleich ihn Mantel und Gewehr beim Laufen hinderten, voraus, um die Bahn entlang zu spähen; war nichts Beunruhigendes in Sicht, so gab er von oben ein Zeichen, die Leute legten sich kräftiger ins Zeug, und
- 59. er selbst kam atemlos wieder herbeigesprungen. Der gefährlichste Punkt war eine Stelle, wo der Zug wieder einen alten Kanaldamm kreuzte: hier hatte die Räuberbande heute Nacht auf der Lauer gelegen. Wir kamen aber unbelästigt auch durch diesen Hohlweg hindurch und an dem Hügel Tell Wabo vorüber. Rechts von der Straße erhob sich in der Ferne Hamudija, eine kleine, an der Karawanenstraße zwischen Bagdad und Hille gelegene Anhöhe. Dann fuhren wir über ein Feld, das mit zahlreichen Ziegelscherben bedeckt war. Welche Schätze aus der Zeit babylonischer Größe mochte wohl diese Erde im Schoße bergen! Über dem Horizont wurden die einfachen Hütten des Dorfes Taldama sichtbar, und vor 6 Uhr waren wir dort. Beim Stationsgebäude lagerte eine Schar Araber auf Strohmatten; sie warteten darauf, ihre Kameraden an den Wagen abzulösen. In der Nähe standen fünfundzwanzig Zelte. Schmale, neuangelegte Kanäle mit kleinen Brücken führten den Feldern Wasser zu, aber offenbar zu wenig, denn das Korn sah kümmerlich aus. Heuschrecken waren hier zahlreich, und viele von ihnen fanden auf den Schienen einen schnellen Tod. Meine Araber liefen, was das Zeug halten wollte; der Schweiß tropfte ihnen von der Stirn, und sie keuchten wie atemlose Hunde. Als einmal zwei leere Wagen die Strecke sperrten, hoben sie das Hindernis einfach vom Gleis herunter und ließen es daneben stehen. Die Sonne war blutrot untergegangen, und die Dämmerungsstunde nahte. Über Bagdad flammten bläuliche Blitze. Am Horizont war der Himmel klar. Die Sterne traten hervor, und der Mond zeigte seine Hörner. Im Norden flackerten die Feuer arabischer Nomaden bei dem Hügel Abu Hanta. Das Stationsgebäude bei Tell-Essued hatte ein auf Pfosten ruhendes Schutzdach, unter dem die Araber lagen und schliefen oder ihre Wasserpfeifen rauchten. Hier wurde Rast gemacht, im „Mangal“, dem Kohlenbecken, Feuer angezündet, und Sale mußte Tee kochen zu einem schnellen Abendessen aus Brot und Eiern. Kurz vor ½9 begann die vierte Wegstrecke.
- 60. Der Mond hatte sich hinter Wolken verkrochen. Über Bagdad aber leuchtete es wie der Widerschein heftigen Artilleriefeuers. Die Gefahr eines Überfalls schien jetzt vorüber zu sein, wenigstens wurde nicht mehr davon gesprochen. Zu beiden Seiten lag die Wüste still und dunkel. Die fünfte Wegstrecke reichte bis Jesr el-Cher. Der Mannschaftswechsel dauerte ein paar Minuten; die alten Leute empfingen ihren Backschisch, und die neuen sahen sich dadurch angespornt, die sechste und letzte Strecke bis zum Tigris mit größter Geschwindigkeit zu nehmen. Auf einer Eisenbahnbrücke überschritten wir den großen Kanal Jesr el-Cher. Dann wieder eine letzte Strecke Feld, und schon tauchten Lichter und Laternen auf, die immer zahlreicher wurden. Palmen traten aus dem Dunkel wie gespenstige Schatten hervor. Nebengleise zweigten sich ab, Güterzüge mit Kriegsmaterial standen hier und dort, und schon hielten wir am Ufer des Tigris. Hammale, Lastträger, kamen gesprungen, bemächtigten sich meines Gepäcks und schleppten es zu einer Treppe, an deren Fuß eine gewaltige Guffa vertäut lag. Wir stiegen an Bord, und drei Mann ergriffen ihre kurzen, breitblattigen Ruder. Sie standen im Vorderteil der Guffa, soweit man bei einem Boot, das wie ein kreisrunder Korb ist, von Vorder- und Hinterteil reden kann, stießen die Ruder mit beiden Händen soweit wie möglich vor dem Boot ins Wasser und arbeiteten sich mit schnellen und immer gleichmäßigen Ruderschlägen vorwärts. Wären nur zwei Ruderer da, so würde das Boot sich bald im Kreise drehen; deshalb arbeitete der mittlere bald mit dem linken, bald mit dem rechten Nachbar, ohne beim Wechseln von der einen Seite zur andern Seite den Takt zu verlieren. So schaukelte das originelle Fahrzeug über den Tigris, Bagdad entgegen. — Bagdad schlief bei meiner Ankunft. Nur hier und da brannte in einem Fenster noch ein Licht oder eine Öllampe. Im übrigen war das linke Tigrisufer stockdunkel. Beim Schein der Blitze waren nur hin und wieder Schattenrisse von Hausdächern, Minaretten und Palmen zu erkennen.
- 61. Wohin nun? Als die Guffa an dem sanft abfallenden Ufer gelandet war, fragte ich die Ruderer, ob sie ein Haus wüßten, wo deutsche Offiziere wohnten. Freilich! Sie schulterten meine Sachen und hießen mich ihnen folgen. Einer von ihnen mußte mich führen, denn Straßenbeleuchtung gab es nicht, und man konnte die Hand nicht vor den Augen sehen; man merkte nur, daß man durch fürchterlichen Schlamm watete. Endlich hielten meine Führer vor einem Tor. Auf dreimaliges Klopfen mit dem Eisenring öffnete ein Diener. „Wer wohnt hier?“ fragte ich. „Einige deutsche Herren. Aber sie sind schon alle zu Bett bis auf einen, der noch nicht zu Haus ist.“ „Ist noch ein Zimmer frei?“ „Ja, eins.“ „Führ mich dahin!“ Vom Hof ging es eine Treppe hinauf über eine offene Galerie in das Zimmer. Ein Licht wurde angezündet, mein Feldbett mit dem Moskitonetz aufgestellt und mein Gepäck auf Tisch und Stühle gelegt. Es war gegen 1 Uhr. Bei wem ich mich einquartiert hatte, ahnte ich nicht. Aber ich machte mir darüber auch keine Gewissensbisse. Denn die Fahrt auf Euphrat und Tigris an einem Tag und auf der merkwürdigen Eisenbahn hatte mich ermüdet, und ich sehnte mich nach Ruhe. Als getreuer Wächter hatte sich Sale vor meiner Zimmertür auf den Boden gelegt. Eben wollte ich unter das Netz kriechen, da erklangen feste Schritte auf der Galerie, und ein deutscher Feldgrauer trat herein. Als er mich erblickte, stutzte er und stand da wie ein fleischgewordenes Fragezeichen. Er hatte das Licht brennen sehen und geglaubt, der rechtmäßige Besitzer des Zimmers sei zurückgekehrt. „Wer ist denn der?“ fragte ich. „Der Tibetforscher Professor Tafel aus Stuttgart.“
- 62. „Wie, Tafel? Mein alter Freund aus der Berliner Richthofen-Zeit!“ „Ja, eben der. Er war einige Tage krank und ist in Behandlung bei Dr. Herle.“ So hatte der Zufall mich, der ich bei stockdunkler Nacht Bagdad betrat, ohne daß jemand von meiner Ankunft wußte, nicht nur in das Haus, sondern auch in das Zimmer geführt, das ausgerechnet ein Tibetforscher bewohnte, und noch dazu einer, den ich schon seit vielen Jahren kannte! Nachdem mich Hauptmann Müller — denn das war der Feldgraue — noch eine Weile über die Verhältnisse in Bagdad unterrichtet hatte, wünschte er mir gute Nacht und überließ mich dem Schlaf.
- 63. Abdurrahaman Gilani, Nakib in Bagdad.
- 64. Das Minarett Suk-el-Gasl in Bagdad. Am Morgen des 6. Mais erwachte ich bei einem wahrhaft tropischen Wetter. Es goß in Strömen; wie Glas stand der Regen vor dem Fenster, er klatschte auf die steifen, blanken Palmenblätter, er schäumte aus den Dachrinnen, rieselte die Veranden herein und brodelte in Strömen über den Hof. Der Donner rollte durch schwere, blauschwarze Wolken. Ohne Zweifel tat das tüchtige Sturzbad der nicht gerade sauberen Stadt recht gut: aber sachkundige Leute meinten, es käme viel zu spät; Regen im Mai sei eine ungewöhnliche Erscheinung.
- 65. Als das Unwetter einigermaßen vorüber war, machte ich mich fertig, durch den Straßenschmutz nach dem Hause des früheren deutschen Konsuls Richarz zu wandern, wo Herzog Adolf Friedrich zu Mecklenburg, mein liebenswürdiger Wirt von Bapaume, seit einiger Zeit sein Zelt aufgeschlagen hatte. Die Wechselfälle des Krieges hatten ihn jetzt nach Asien geführt; sein Wunsch, als Führer einer eigenen kleinen Armeegruppe an einer der türkischen Fronten kämpfen zu dürfen, war jedoch zu seinem Bedauern nicht in Erfüllung gegangen. Übrigens war er kein Neuling in diesem Lande, denn in seiner Jugend, vor ungefähr zwanzig Jahren, war er von Jerusalem nach Berlin geritten. Eben trat ich aus meinem Zimmer auf die Galerie hinaus, da kam mir der Herzog schon entgegen, frisch und munter wie gewöhnlich. Er hatte von meinem nächtlichen Einzug gehört und wollte der erste sein, der mich willkommen hieß. In unsere Regenmäntel gehüllt wanderten wir über die vornehmste Straße Bagdads, die Halil Paschas Namen trägt, zum Hause des Herrn Richarz, einem gewaltigen Viereck, das ein schattenspendender Garten umgab. Offene, überdachte Galerien, die auf geschnitzten und buntbemalten, aber verwitterten Säulen ruhten, boten freie Aussicht über Nachbarhöfe und enge Gassen, auf den großen königlichen Strom und sein anderes Ufer. Herr Richarz war 1894 bis 1907 deutscher Konsul in Bagdad und von 1912 bis 1914 amerikanischer. Dann hatte er seinen Abschied genommen, war aber hier wohnen geblieben; er mochte sein Haus nicht im Stich lassen, es auch nicht während des Krieges zu einem Spottpreis verkaufen. Außerdem liebte er Bagdad und hatte sich im Lauf der Jahre an sein eigenartiges Klima gewöhnt und sich in diese bunte orientalische Welt, ihre Sitten und Sprachen — Richarz beherrschte fließend ihrer elf — so eingelebt, daß er mit seinen sechzig Jahren dieses ruhige, sorgenlose Dasein nicht ohne Zwang aufgeben wollte. Wie dunkel und farblos waren die Straßen Berlins und Hamburgs verglichen mit den Gassen Bagdads! Mit der Genauigkeit eines Uhrwerkes hatte er sein Tagewerk geregelt; pünktlich zur Sekunde nahm er seine Mahlzeiten, seine Bäder, machte er seine Spaziergänge, las er seine stark verspätet
- 66. einlaufenden Zeitungen oder die Schätze seiner Bibliothek, und ebenso regelmäßig verrichtete er seine Arbeiten für den deutschen Nachrichtendienst. Salon, Arbeitszimmer, Bibliothek und Speisesaal gingen auf die Galerie hinaus, die um den Hof lief. Im Salon stand ein über Meer und Ströme beförderter Flügel, dessen Innerm sein Besitzer schöne Melodien entlockte, denn er war sehr musikalisch und fand in einsamen Stunden am Klavier die beste Gesellschaft. Graf Wilamowitz und Konsul Richarz auf dessen Kai in Bagdad
- 67. Welcome to our website – the ideal destination for book lovers and knowledge seekers. With a mission to inspire endlessly, we offer a vast collection of books, ranging from classic literary works to specialized publications, self-development books, and children's literature. Each book is a new journey of discovery, expanding knowledge and enriching the soul of the reade Our website is not just a platform for buying books, but a bridge connecting readers to the timeless values of culture and wisdom. With an elegant, user-friendly interface and an intelligent search system, we are committed to providing a quick and convenient shopping experience. Additionally, our special promotions and home delivery services ensure that you save time and fully enjoy the joy of reading. Let us accompany you on the journey of exploring knowledge and personal growth! ebookfinal.com