Understanding Unexpected Failure Scenarios from Aircraft Systems Virtual Test Data

Current goals for emission-free aviation require novel system concepts, such as fuel cell-driven propulsion. Yet, such concepts do not exist in current aircraft. With such low operational experience, the requirements for system design may not be fully understood. Consequently, there is a rising potential for “blind spots” in the design that emerge in integration testing or even later, leading to costly adaptations. These unexpected scenarios may be discovered early by intelligent exploration during virtual, simulation-based testing. However, there is still a gap between the physical-dynamic test data and the functional-logic requirements definition by engineers. Test results may contain many data measurements. For a human engineer, it is time-consuming and error-prone to process the amount of time series data, making it impractical in an industrial setting. To enable the engineering evaluation and derivation of missing requirements, a computer-aided abstraction step is needed. This work presents a framework to derive principal functional-logic scenarios from critical test data and to present them visually. Concretely, analysis agents extend the data with discrete system states, followed by feature-based clustering. Finally, single sequences are derived for each cluster. The results are visualized as functional-logic parallel lifeline charts. The approach is evaluated using a fuel cell-driven propulsion example.