A BAYESIAN APPROACH TO SOLVING PROBLEMS OF DIAGNOSTICS OF ELECTRICAL EQUIPMENT IN ROBOTIC SYSTEMS

Abstract

Background. Given the increasing complexity of electrical equipment in robotic systems, a reliable assessment of their technical condition is particularly important. Traditional methods of periodic monitoring do not allow for the sufficient consideration of accumulated diagnostic information and adequate assessment of the risks of erroneous decisions. The aim of this study is to substantiate a Bayesian approach to processing diagnostic information obtained through technical condition monitoring, followed by the development of a reliable diagnosis. Materials and methods. A two-stage algorithm for statistical processing of diagnostic information is proposed. At the first stage, posterior probabilities of condition membership are estimated using Bayes' theorem and kernel density estimation. At the second stage, a decision is made based on minimizing the average risk, allowing for the cost of classification errors to be taken into account. The zone method structures the condition space. Diagnostic features are divided into intervals, which serve as the basis for identifying condition classes. Results. A two-stage algorithm for deciding on the type of condition has been developed, combining probabilistic assessment and optimization based on the minimum risk criterion. Conclusions. Condition monitoring combined with a Bayesian approach to processing diagnostic information ensures higher reliability in determining the condition type of electrical equipment in robotic systems compared to traditional periodic monitoring methods.