Computational design process modelling

Abstract

In the conceptual design phase, relatively simple equations and functions (or compiled code) are used to describe the aircraft and to perform trade-off studies. The latter require an optimal execution sequence in order to reduce computational cost and design time, respectively. The focus of this paper is the dynamic derivation of the optimal computational plan for each study so that the designer could focus on designing the aircraft rather than managing the process flow. Two methodologies, the Design Structure Matrix (DSM) and the Incidence Matrix are used for the computational process modeling. The incidence matrix describes the relationship between variables and equations/models. The DSM has been used to express the dependency relationships between the models and also, after manipulation, to produce the solution process. The designer specifies the independent (known) variables first. Then the variable flow is modeled using the Incidence Matrix Method (IMM). It determines how data flows through the models, and also identifies any strongly connected components (SCCs). The second step is to rearrange all equations/models hierarchically in order to reduce the feedback loops in each of the identified SCCs. This is achieved by the application of a genetic-based algorithm. Subsequently all SCCs and noncoupled models are assembled into a macro model which forms a global DSM. The global DSM is further rearranged to obtain an upper triangular matrix which defines the final model execution sequence. A simple aircraft sizing example is presented to illustrate the proposed method and algorithm. Advantages of the method include improved efficiency and the ability to deal with both algebraic and numerical models as well as with multiple outputs per model.