Formal Design of Fault Detection and Identification Components Using Temporal Epistemic Logic

Automated detection of faults and timely recovery are fundamental features for autonomous critical systems. Fault Detection and Identification (FDI) components are designed to detect faults on-board, by reading data from sensors and triggering predefined alarms. The design of effective FDI components is an extremely hard problem, also due to the lack of a complete theoretical foundation, and of precise specification and validation techniques. In this paper, we present the first formal framework for the design of FDI for discrete event systems. We propose a logical language for the specification of FDI requirements that accounts for a wide class of practical requirements, including novel aspects such as maximality and non-diagnosability. The language is equipped with a clear semantics based on temporal epistemic logic. We discuss how to validate the requirements and how to verify that a given FDI component satisfies them. Finally, we develop an algorithm for the synthesis of correct-by-construction FDI components, and report on the applicability of the framework on an industrial case-study coming from aerospace.