Diagnosis of HV stator bars insulation in the presence of multi partial-discharge phenomena

Results of Partial Discharge (PD) measurements, performed on insulation system of rotating machines, are processed referring to the Weibull statistics. Probability functions based on the additive-mixed Weibull distribution are used to fit experimental data obtained in the presence of PD phenomena of different types, simultaneously active. A non-parametric Kernel Regression technique is adopted to pre-process the experimental data in order to determine the number of components and the basic characteristics of the mixed distribution. A tree structure ibr the PD phenomena clustering particularly useful to monitor the time and/or voltage behavior of multi-PD phenomena, is proposed.