| description abstract | This paper presents a mathematical model for quantifying uncertainty of a discrete design solution and to monitor it through the design process. In the presented entropic view, uncertainty is highest at the beginning of the process as little information is known about the solution. As additional information is acquired or generated, the solution becomes increasingly well-defined and uncertainty reduces, finally diminishing to zero at the end of the process when the design is fully defined. In previous research, three components of design complexity—size, coupling, and solvability—were identified. In this research, these metrics are used to model solution uncertainty based on the search spaces of the variables (size) and the compatibility between variable values (coupling). Solvability of the variables is assumed uniform for simplicity. Design decisions are modeled as choosing a value, or a reduced set of values, from the existing search space of a variable, thus, reducing its uncertainty. Coupling is measured as the reduction of a variable’s search space as an effect of reducing the search space of another variable. This model is then used to monitor uncertainty reduction through a design process, leading to three strategies that prescribe deciding the variables in the order of their uncertainty, number of dependents, or the influence of on other variables. Comparison between these strategies shows how size and coupling of variables in a design can be used to determine task sequencing strategy for fast design convergence. | |
