Efficient Population of the Verification Tradespace Using Bayesian Inference

Abstract

This article presents a process to efficiently elicit belief models of verification strategies in the context of applying tradespace exploration to the design of verification strategies. The need for efficiency is at the core of tradespace exploration, where a large region of the solution space needs to be explored before anchoring to a specific solution or solution set. In conceptual design and systems architecture, solutions are modeled by leveraging independent models that are connected by an overarching model. In designing verification strategies, however, verification activities cannot be modeled independently of the overall strategy in which they are used. As a result, probabilities associated with the confidence generated by each verification activity, need to be elicited for each candidate verification strategy individually. This requires, in principle, to generate a specific model for each solution in the solution space, which may hamper efficiency. In this article, we show that Bayesian inference can be used to overcome this limitation. In particular, the proposed process in this article uses only one elicitation activity for an overarching verification strategy and then leverages Bayesian inference to determine automatically the probabilities of other verification strategies in the solution space.