New results on discrete-event counting under reliable and unreliable observation information

We present an approach for addressing the issues of detecting repeated fault events in the framework of model-based monitoring of discrete-event systems (DES) under reliable and unreliable observation information. The analysis task is to determine whether a certain observation configuration is capable of reporting the occurrence of fault events while satisfying the performance requirements. If the reliability of observation information is assured, the assessment is accomplished by evaluating diagnosability notions of interest. To evaluate the notions of diagnosability regarding repeated fault counting, polynomial-time verification algorithms are developed. In order to deal with unreliable observation information, the concept of detection confidence is introduced, which measures the quality of fault counting. An algorithm computing detection confidence is conjectured. For online fault counting, we develop a new online fault counting algorithm assuming observation reliability. The developed algorithm has lower time and space complexities than an online diagnosis algorithm reported in literature for counting the occurrence of repeated faults. This online algorithm is naturally extended to handle the unreliable observation information.