Statistic Based Method for Post-processing Analysis in Lifetime Investigations of Multi-factor Aged Winding Insulation

Abstract

To evaluate the condition of electrical insulation systems the knowledge of predominant aging mechanisms is of particular importance. As experiments are very expensive and require a lot of time, and the experimental setup only permits a limited number of test samples at the same time, it is essential to analyze the already available lifetime data with the use of statistical methods within post-processing. Rotating machines are important components in power generation. There is a high level of demand for machine reliability and availability. Negative influences on machine reliability and life expectancy might be thermal, electrical and mechanical stress. Integration of decentralized renewable energy sources lead to a fluctuating load profile of power engineering equipment, such as generators. This increasingly stresses the electrical insulation system in form of mechanical and thermomechanical forces. Regarding mechanical aging the condition of winding insulation of large rotating machines has to be investigated more detailed. In this context, experimental investigations with a focus on mechanical ageing are necessary to understand the complex insulation system in order to prevent failures. Previously, initial evaluations using Weibull distribution have been used. It has to be noted that only a small sample size is available, which limits a statistical significant evaluation. For this reason, different statistic methods are used to optimize the empirical data post-processing for assessing statistical accuracy and to improve the quality of predictions for small sample size. Using multiple regression and bootstrapping, present empirical data are analyzed.