The Fault-Augmented Approach for the Systematic Simulation of Fault Behavior in Multi-Domain Systems in Aerospace 2018-01-1917

A library for modelling faults in multi-domain physical systems is introduced. The library is based on the simulation of fault effects on the system’s behavior. The motivation of how and why to model faults systematically as well as a description of the Modelica®-based library structure with a wizard supporting the semi-automatic augmentation process of faults are outlined. The fault types are classified into continuous and discrete with dedicated type definitions. The application of the Fault library is exemplified in the field of aerospace electrohydraulic actuator. The actuator is equipped with hydromechanical, electrical and digital systems for mitigating failures, which should be tested at an early stage of design. To perform the tests, a multi-domain, dynamic system model is created, wherein failures are systematically simulated using a special approach for fault augmentation. In addition, several complementary tests are obtained by a variants simulation and the simulation results of the fault augmented model are analyzed and using supervised machine learning classifier are demonstrated. Different classification algorithms were compared to each other and analyzed. The accuracy of an appropriate machine learning classifier is analyzed in detail to classify several faults from different domains and to localize their impact by changing a control mode. The selection of relevant output values for the fault classification in the electrohydraulic control system is executed based on the extraction of the feature’s importance.