Probabilistic extension of failure net based FMEA

Failure Mode and Effects Analysis (FMEA) is one of the most widely used techniques for identifying risks in complex technical systems. It is a mandatory practice in most safety-critical domains. Initially, FMEA was a tabular technique, listing potential failures with their possible effects, thereby suggesting a cause-effect relationship. In later years, it had been extended towards a graphical technique enabling a display of the cause-effect relationships across several hierarchy levels in the form of failure nets (FN). Later approaches and FMEA software tools suggest the use of logical operations to model OR and AND combinations of single and multiple failures in FN, but an underlying formalism is still missing. Also, there is no general consensus whether FN are suited to represent the actual causality between failure events or just represent logical conjunctions of fault states. Moreover, as there is no feature to weight FNedges, it is not possible to express the likelihood of potential failure consequences, which may result in overly conservative estimates. A potential solution is to assign conditional probabilities to the edges of FN. Next, it is not straightforward to model the functional interaction of mechanisms at operation time which prevent cause-effect-chains from evolving, for example by safety-related functions that counteract the failure consequences. Such detection mechanisms may bring the system into a safe state and potentially cover several faults originating from different components, thereby influencing the probability of the failure consequences. But the application of FMEA to systems in which safety measures typically are implemented, e.g. software-determined systems, is not sufficiently solved yet. At last, the restricted compatibility with other techniques as the FMEDA or FTA keeps databases and cognition in separate slots, which increases the risk of inconsistencies and hampers the integration of analyses from different suppliers. To overcome these drawbacks, we propose a semantic underpinning to FMEA that (1) includes fault conditions and failure events, (2) reflects the propagation of failures between components at different architecture hierarchy levels, (3) refers to the causal meaning of FN-edges, (4) allows for the assignment of conditional probabilities to these edges and (5) provides extended logical combinations for FN along with a specific notation. We present appropriate calculation laws and show the compatibility to common rules of related techniques such as Fault Trees and FMEDA. The application of this probabilistically extended FMEA technique is demonstrated by a practical example taking the modeling of safety-mechanisms into account