Computer-implemented method for generating a component fault and deficiency tree of a multi-component system comprising a plurality of components

A computer-implemented method for generating a Component Fault and Deficiency Tree of a machine learning-based system with safety-critical functionalities comprising a plurality of components, wherein

• each component of the plurality of components of the machine learning-based system is a hardware component, a software component, or a hardware and software component,
• the method comprising:
• modeling the machine learning-based system using a Component Fault and Deficiency Tree,
• wherein the Component Fault and Deficiency Tree comprises a plurality of component fault and deficiency tree elements associated with the respective components of the machine learning-based system;
• and
• interconnections between the component fault and deficiency tree elements associated with respective functional dependencies between the components;
• wherein each component fault and deficiency tree element comprises at least one inport and at least one outport;
• wherein each component fault and deficiency tree element comprises at least one output failure mode, connected to the at least one outport;
• wherein each component fault and deficiency tree element comprises at least one input failure mode, connected to the at least one inport;
• wherein each component fault and deficiency tree element comprises at least two events as internal fault tree logic;
• wherein one event of the at least two events is associated with at least one failure and the other event of the at least two events is associated with at least one deficiency of the component;
• wherein at least one gate is configured to connect the at least two events;
• wherein each component fault and deficiency tree element comprises at least one mitigation logic;
• wherein the at least one mitigation logic is configured to mitigate the at least one failure and at least one deficiency by means of a mitigation measure;
• wherein at least one Boolean AND-Gate is configured to connect the internal fault tree logic and the at least one mitigation logic;
• and
• providing the Component Fault and Deficiency Tree of machine learning-based system as output;
• and
• applying at least one safety analysis technique on the machine learning-based system.