Component-based combination of online-diagnosis methods using diagnostic directed acyclic graphs

In safety-critical application domains online fault-diagnosis contributes to a significantly increased system reliability and safety by detecting and diagnosing occurred faults and, if applicable, making the system recover from faults. In order to enable fault-specific recovery actions, e.g., a reconfiguration of the system, cause-based fault identification is needed. This typically requires a large amount of data to be analyzed and evaluated during a diagnostic process. For sound decisions on occurred faults within complex systems, it is often beneficial to combine several online-diagnosis methods. In this work we present a component-based diagnostic framework based on a diagnostic dependency graph. Herein, multiple online-diagnosis methods are combined in the form of encapsulated tasks. The dependencies as well as the tasks adapt to the system at run time. Our diagnostic framework is demonstrated by means of an automotive use-case.