Estimation of the residual life of steam-turbine condensers based on statistical models

The article presents the results of a study aimed at estimation of the state and prediction of the residual life of steam-turbine condensers on the basis of statistical analysis. The possibility of such evaluation during the operation of steam-turbine plants with accuracy sufficient for practical purposes is demonstrated. It is shown that identification of the operating period, viz., the initial period, the normal operation period, or the period of the lifetime exhaustion, as well as determination of the condenser’s operating time at the moment when the failure of an individual tube occurs, is very important for statistical evaluation of the condenser state. Two statistical models are proposed and comparative analysis of the results calculated by these models for the residual life of the condensers at the Reftinskaya SDPP has been performed. The first model can be used when comprehensive information about the condenser’s operating time before the tubes have failed is available as well as a priori information—or information based on analysis of the condenser tube metal—that the condenser is in its normal operation period. In this case, the fact of exhaustion of the condenser’s lifetime is established by reaching the limit of the failed condenser tubes, which is determined by technical and economic analysis of losses caused by operating the turbine with a reduced heat-exchange surface of the condenser. The distribution function for the operating time of the failed tubes is approximated by a normal distribution. In the cases when no precise information on the condenser tubes’ operating time is available at the thermoelectric power plant (TEPP), the second statistical model based on censored samples is proposed for estimation of the condenser state. An expression to assess the confidence interval that determines the significant difference between the distribution functions for complete and censored operating time values has been derived. It is shown that this model allows prediction of the beginning of the condenser’s lifetime exhaustion period that is 1.5–2.0 years distant from the moment of mass failure occurrences.