Maintaining consistency in the design process

摘要

The typical design process follows a sequential though iterative method. A design naturally moves sequentially through design stages from concept through realization. There are numerous aspects of design, however, that require a non-sequential iteration in the design process, including: the high availability of new information, inadequate modeling of real-world interactions, concurrent engineering issues, and the existence of design requirements that the design system cannot directly input. Non-sequential design events are a common occurrence in the design process and must be incorporated into any design model. In this thesis, the use of iteration in the design process is studied and a design paradigm, diagonalized design, is discussed that integrates the sequential and iterative natures of design. Design consistency is then formalized and presented as a critical component of any diagonalized design system. Numerical consistency techniques are then developed, proven, and implemented in two different design areas.;First, a variational mechanical design system is demonstrated where the dimensions of a part are constrained with nonlinear equations representing high-order attributes of the part (e.g., mass, surface area, and maximum load). As systems of nonlinear equations typically have multiple solutions, an interval-based continuation method is developed (COAST) to follow a specific solution as the constraints are modified. Examples shown in the thesis include the design of a cantilever beam, worm gear, and helical spring. Constraint-based design of faired parametric curves is then demonstrated when the constraints are on the whole curve (e.g., a curve having a certain arc length or being a given distance from another graphical object) and not solely on the defining points of the curve. Because the constraints will often be changed over time, it is shown that finding the global optimum of the fairing objective function is often less important than finding a consistent solution and that a local optimum of the objective function should at least be an option in such a system. Necessary conditions of the optimization problem are used to locate the desired local optimum. As there are normally many local optimums, the COAST methodology is applied to maintain a consistent solution. Besides demonstrating this technique with relations between Bezier curves, an example is shown from apparel design.;Finally, as a proof-of-concept, a 3-D shoe design system is demonstrated that can grade an upper design to any size by modifying the measurements of the parametrized model of a foot. The system is then able to create the corresponding manufacturable patterns and render the design.