Abstract

As digital systems grow in complexity and are used in a broader variety of safety-critical applications, there is an ever-increasing demand for assessing the dependability and safety of such systems, especially when subjected to hazardous environments. As a result, it is important to identify and correct any functional abnormalities and component faults as early as possible in order to minimize performance degradation and to avoid potential perilous situations. Existing techniques often lack the capacity to perform a comprehensive
and exhaustive analysis on complex redundant architectures, leading to less than optimal risk evaluation. Hence, an early analysis of dependability of such safety-critical applications enables designers to develop systems that meets high dependability requirements. Existing techniques in the field often lack the capacity to perform full system analyses due to state-explosion limitations (such as transistor and gate-level analyses), or due to the time and monetary costs attached to them (such as simulation, emulation, and physical testing).

In this work we develop a system-level methodology to model and analyze the effects of Single Event Upsets (SEUs) in cyberphysical system designs. The proposed methodology investigates the impacts of SEUs in the entire system model (fault tree level), including SEU propagation paths, logical masking of errors, vulnerability to specific events, and critical nodes. The methodology also provides insights on a system's weaknesses, such as the impact of each component to the system's vulnerability, as well as hidden sources of failure, such as latent faults. Moreover, the proposed methodology is able to identify and categorize the system's components in order of criticality, and to evaluate different approaches to the mitigation of such criticality (in the form of different configurations of TMR) in order to obtain the most efficient mitigation solution available.

The proposed methodology is also able to model and analyze system components individually (system component level), in order to more accurately estimate the component's vulnerability to SEUs. In this case, a more refined analysis of the component is conducted, which enables us to identify the source of the component's criticality. Thereafter, a second mitigation mechanic (internal to the component) takes place, in order to evaluate the gains and costs of applying different configurations of TMR to the component internally. Finally, our approach will draw a comparison between the results obtained at both levels of analysis in order to evaluate the most efficient way of improving the targeted system design.