Combining software quality analysis with dynamic event/fault trees for high assurance systems engineering

We present a novel approach for probabilistic risk assessment (PRA) of systems which require high assurance that they will function as intended. Our approach uses a new model i.e., a dynamic event/fault tree (DEFT) as a graphical and logical method to reason about and identify dependencies between system components, software components, failure events and system outcome modes. The method also explicitly includes software in the analysis and quantifies the contribution of the software components to overall system risk/ reliability. The latter is performed via software quality analysis (SQA) where we use a Bayesian network (BN) model that includes diverse sources of evidence about fault introduction into software; specifically, information from the software development process and product metrics. We illustrate our approach by applying it to the propulsion system, of the miniature autonomous extravehicular robotic camera (mini-AERCam). The software component considered fo r the analysis is the related guidance, navigation and control (GN&C) component. The results of SQA indicate a close correspondence between the BN model estimates and the developer estimates of software defect content. These results are then used in an existing theory of worst-case reliability to quantify the basic event probability of the software component in the DEFT