Aircraft design system requirements analysis with Bayesian Networks

This paper introduces readers with the application of Bayesian Networks in conceptual design tasks. These models can represent in a graph the cause-and-effect logic of requirements and enable the analysis of uncertainty in decision making. A syntax for converting a requirement into a Bayesian network query is also presented. These concepts are applied on a wing design example to demonstrate the flexibility of a Bayesian Network in answering design-related questions. In particular, a feasible design space is identified through inference of the requirements onto the input parameters. This is useful to narrow down configurations for further detailed analysis with an optimization algorithm. The model is capable of estimating the quantities of interest and the effect of the requirement uncertainty on the system variables. Through proper data collection and integration with digital engineering methods, a system-wide Bayesian network would enable reasoning between systems and components, accelerating the industrial development of complex products.