Model-Based Systems Engineering with Machine Learning for Test Case Generation

In complex aerospace systems, unexpected combinations of multiple parameters can lead to surprising systemic failures. In large parameter spaces, such as for urban air mobility (UAM) applications, it can be difficult to determine which combinations of parameters were key to driving an unwanted outcome. To address these limitations, we propose developing and applying a Model-Based Systems Engineering (MBSE) solution called Identifying Parameters Leading to Requirements Violations (IPLRV) to a complex UAM aircraft model to identify the combination of boundary parameters that could lead to accidents. MBSE is an effective methodology for developing requirements, system design and performing early verification and validation of complex systems models. The proposed IPLRV is a fast, effective toolchain used to determine key parameter ranges that lead to failure of requirements for complex black-box aviation models. The four main components of the IPLRV are a Data Dictionary and Electronic Interface Control Document that enables easy substitution of system models, a GUI for simple specification of requirements and derived data, integration of a parameter learning technique which quickly prunes dimensions and drives the simulation towards key parameter ranges, and integrated graphics and simplified reporting of parameter ranges.