Design is a difficult and complex process requiring; creativity, experience, domain knowledge, and problem solving skills. Much of the information that is used and generated during the design process is rarely explicitly recorded. This includes the reasons why design decisions were made. This information is commonly referred to as design rationale (DR). As a result many of the tasks that are performed during the design process are still poorly understood and modifications to designs can have unforeseen and possibly dangerous consequences.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

Constraint-based systems are increasingly being used to support the design of products. Several commercial design systems based on constraints allow the geometry of a product to be specified and modified in a more natural and efficient way. However, it is now widely recognised the needs to have a close coupling of geometric constraints (i.e. parallel, tangent, etc) and engineering constraints (Le. performance, costs, weight, etc) to effectively support the preliminary design stages. This is an active research topic which is the subject of this thesis. As the design evolves, the size of the quation set which captures the constraints can get very large depending on the complexity of the product being designed. These constraints are expected to be solved efficiently to guarantee immediate feedback to the designer. Such requirement is also necessary to support constraint-based design within Virtual Environments, where it is necessary to have interactive speed. However, the majority of constraint-based design systems re-satisfy all constraints from scratch after the insertion of a new design constraint. This process is time consuming and therefore hinders interactive design performance when dealing with large constraint sets. This thesis reports research into the investigation of techniques to support interactive constraint-based design. The main focus of this work is on the development of incremental graph-based algorithms for satisfying a coupled set of engineering and geometric constraints. In this research, the design constraints, represented as simultaneous sets of linear and non-linear equations, are stored in a directed graph called Equation Graph. When a new constraint is imposed, local constraint propagation techniques are used to satisfy the constraint and update the current graph solution, incrementally. Constraint cycles are locally identified and satisfied within the Equation Graph. Therefore, these algorithms efiiciently solve large constraint sets to support interactive design. Techniques to support under-constrained geometry are also considered in this research. The concept of soft constraints is introduced to represent the degrees of freedom of the geometric entities. This is used to allow the incremental satisfaction of newly imposed constraints by exploiting under-constrained space. These soft constraints are also used to support direct manipulation of under-constrained geometric entities, enabling the designers to test the kinematic behaviour of the current assembly. A prototype constraint-based design system has been developed to demonstrate the feasibility of these algorithms to support preliminary design.

Communication between different members of a design team often poses difficulties. The knitwear design process is shared by the designers, who plan the visual and tactile appearance of the garments, and the technicians, who have to realise the garment on a knitting machine and assemble it. This thesis reports a detailed empirical study of over' twenty companies in Britain and Germany, which shows that the communication problem constitutes a major bottleneck. Designers specify their designs inaccurately, incompletely and inconsistently; the technicians interpret these specifications according to their previous experience of similar designs, and produce garments very different from the designers' original intention. Knitwear is inherently difficult to describe, as no simple and complete notation exists for knitted structures; and the relationship between visual appearance and structure and technical properties of knitted fabric is subtle and complex. At the same time the interaction between designers and technicians is badly managed in many companies. This thesis argues that this communication bottleneck can be overcome by enabling designers to produce accurate specifications of technically correct designs, through the help. of an intelligent computer support system that corrects inconsistent input and proposes design suggestions that the user can edit. In this thesis this proposal is elaborated for one aspect of knitwear design: garment shape construction. Garment shapes are modelled using Bezier curves generated using design heuristics drawn from industrial practice, to create curves that look right to a designer and can be easily edited. The development of the garment shape models presented in this thesis involved the solution of unusual problems in numerical analysis. The thesis shows how the mathematical models can be integrated into an intelligent CAD system, and discusses die benefits of such a system could have for the design process.

This thesis describes a formal methodology for defining and assessing product performance and its implementation in a prototype computer system. The methodology is based on abstract descriptions of the operations that are conducted within the design process. It is, consequently, extremely generic and creates a bridge between physical product performance and actual user requirements. The methodology is based on defining product attributes in terms of observable parameters of the product in use. Defining an attribute in this way inherently reflects its required interaction with the user and consequently can truly be said to be in "user terms" A product will have a range of attributes and a performance indicator is proposed, such that the attributes are combined in a way that reflects their relative importance to the user. At the conceptual stage of the design process, when the actual product does not exist, and only some abstract representation is available, it is vitally important to be able to model or simulate and hence evaluate the product attributes. This area of design has often been associated with non algorithmic design procedures, because of its intangible nature. In this thesis the attribute methodology has been used to implement a prototype Computer Aided Design Evaluation Tool (CADET), which has been used and tested with an existing product range. The methodology being abstractly defined supports a wide range of product attributes. It also gives an indication of how the correspondingly wide range of existing analysis software could be integrated into a powerful single Computer Aided Design system. This work has resulted in the publication of two papers in refereed Journals and the presentation of eight other papers at refereed International Conferences. A list of the publications is included in the Appendices.