Artificial intelligence applied to the aging of machine insulation

The intense interest in applying diagnostic techniques to large electrical machines has centered, understandably, on the use of partial discharge (PD) parameters which can readily be measured on line. However, attempts to apply a variety of artificial intelligence methods to the partial discharge data have met with very mixed success. This paper explores the possibility of using a combination of electrical and acoustic diagnostic measurements for the purpose of estimating the state of the insulation system. An automated intelligent system is presented which has shown itself capable of identifying the condition of a set of sample stator bars. In this way, both conventional and derived partial discharge signatures are used in combination with measurements taken from an active (pitch-catch) acoustic probe. The latter technique contributes to the method by permitting accurate identification of delamination and void content assessment. The intelligent system developed permits the identification of four different stages in the process of deterioration. In each of these stages, the weighting of the contributing signatures is adjusted based on the sensitivity of the parameter