This paper describes the key problems in the tasks of intelligent patterns recognition for the state of power circuit breakers and the ways to solve them. The paper presents results of approbation of the system for assessing the technical state of power circuit breakers developed by the authors based on machine learning methods to optimize the schedules of circuit breakers repair and replacement based on analysis of technical diagnostics data and retrospective data. The main requirements for the training sample for recognizing the patterns of the circuit breakers state were also revealed.