View article

The paper deals with the problem of mathematical models building for diagnostic variables of radio electronic systems in the case of data heteroskedasticity. The heteroskedasticity is associated with event when clusters from single sample are characterized by different values of variance. In operation theory heteroskedasticity occurs in case of equipment technical condition deterioration. The paper describes the method for taking into account heteroskedasticity, which includes the following steps: building and choosing the best regression model; calculation of heteroskedasticity index based on minimization of weighted sum of residuals squares; determining correction coefficients of heteroskedasticity and obtaining the final regression equation. The proposed methods can be used during utilization of condition-based maintenance with diagnostic variables monitoring.