Faculté de l'environnement naturel, architectural et construit ENAC, Département de génie civil, Institut de structures IS (Laboratoire de la construction métallique ICOM)

Preliminary bridge design navigation tool for novices

Boulanger, Sylvie ; Hirt, Manfred A. (Dir.)

Thèse Ecole polytechnique fédérale de Lausanne EPFL : 1997 ; no 1735.

Add to personal list
    The motivation of the thesis comes from the frustrations of young engineers confronted with real design problems. The inspiration of the thesis evolved from observations of bridge designers and analyses of bridge design competitions. Not only do designers adopt more than one strategy during design, they rarely perform a fixed sequence of tasks. Not only do designers consider more than one criterion during design, their priorities shift during the determination of parameters. The choice of task sequence, which is largely reflected in the style of the designer, is considered a major source of design variety. The thesis is founded on the premise that design is path dependent. Hence, freedom of sequence and support for an opportunistic strategy, are essential aspects to acquire, model and implement, in order to capture realistic design situations, to achieve design variety for the same initial conditions and to apply these aspects to interactive computational environments. In this thesis, an integrated approach was adopted throughout acquisition, modelling and implementation, in order to propose formalisms, a framework and a design tool that supports freedom of sequence. The formalisms consist of four design characteristics (tasks, parameters, strategies and criteria), a decision making process (the labyrinth metaphor) and a control process (design navigation). The style of the designer is derived as the strategies and criteria adopted for making decisions about tasks and parameters. The labyrinth metaphor is considered the most realistic reflection of decision making paths observed with designers. Design navigation highlights a designer's skill for controlling the labyrinth, without getting lost. However, more acquisition is needed to strengthen the strategies proposed. The framework integrates these formalisms into four parts: a design space, a workspace, a decision path and design interactions. The notion of a workspace, as a continually changing subset of the design space, prepares the labyrinth metaphor for implementation into the interface. Design interactions model the interactions between the design space and the workspace, for navigating in the labyrinth of tasks and parameters. However, a more extensive design space is still required for improving the framework's functionalities, in particular for interacting with CAD tools. The design tool's interface is an implementation of the framework. The interface rests on two dynamic window environments called the task workspace and the parameter workspace. Each workspace has a labyrinth area and a navigation area, which separates decision making from control, explicitly. A prototype of the design tool, called FIBRES, was demonstrated to practitioners and students, whose feedback confirmed the importance of allowing the user freedom of sequence and control over interactions. However, a more robust prototype is needed to perform extensive testing with students. A user-interface that accommodates unrestricted task sequencing has the potential to train novice designers in good bridge design by inducing an implicit mode of learning, and to provide experienced designers with computer workspaces compatible with their own decision making style. In the long term, such an approach may contribute to a better acceptance and penetration of artificial intelligence techniques into civil engineering education and practice.