Application of binary classifier metrics for constructing assessments of the reliability of monitoring the complex technical systems fleet state

The problem of constructing estimates of the reliability of monitoring the technical condition of a fleet of complex technical systems (CTS) with metrological support is investigated. The classical semi-Markov stationary model of CTS functioning used includes the following technical conditions: operational, failure, verification of a failed CTS sample, verification of a working sample, false failure condition, undetected failure condition, restoration (repair). The main indicator of the effectiveness of the CTS fleet is the coefficient of readiness for its intended use. To maintain the availability coefficient at a high level, periodic inspections are carried out during normal operation of the fleet with optimal frequency and optimal tolerance for the controlled parameter. To assess the reliability of finding the CTS in these states, four indicators are proposed: correctness, accuracy, completeness, and the F1 metric, which are used in the theory of artificial intelligence as binary classifier metrics. The simulation of the processes of the CTS fleet functioning in different modes and under different conditions has been carried out. The effect of the interval between verifications, tolerance on the controlled parameter, failure rate, recovery time on reliability indicators and the coefficient is shown ready. The use of reliability indicators will allow the operator of the CTS park with metrological support to comprehensively analyze up-to-date operational information about the current state of the park, develop and apply park management strategies depending on the current situation. The results obtained can be implemented and used in the decision support system for operational remote control of the CTS fleet.