Learning Temporal Truth Tables of Dynamic Fault Trees from Time Series Data on Faults
Parisa Niloofar, Sanja Lazarova‐Molnar · 2023
Classical fault tree analysis (FTA) can be used to analyze and assess combinations of failures of basic components that lead to system failures within a given system. Classical FTA is however unable to handle how temporal sequences of faults lead to system failures. This is an important issue in complex dynamic systems. To model complex systems where the behavior of components or events depends on the system's state or time, dynamic gates are used. Therefore, we extend our work on learning static repairable fault trees from time series data and propose a data-driven algorithm to extract dynamic gates from time series data of faults. Our proposed algorithm captures sequences of events by converting a time-stamped truth table into temporal gates, which allows learning dynamic gates from data.