Industrial electrical devices are designed in a way so that they can function for decades. The insulation used in such devices must therefore meet a number of quality requirements. Under working conditions, the insulating materials are subjected to different electrical, mechanical and thermal effects. These effects damage the insulations and can create cavities inside them, wherein partial discharges (PD) can occur. Partial discharges are small electrical discharges inside or on the surface of materials. PDs further damage the material which eventually leads to the failure of the insulation and possibly the whole device. For this reason it is essential to recognize, classify and measure them. With electrical measurements it is possible to obtain information about the condition of the isolation and the likelihood of its failure. These measurements provide valuable data which can be used to approximate the required maintenance costs and times. At the beginning of my thesis I give a detailed description about the physical background of the partial discharge phenomenon and present several isolation diagnostic measurement techniques. Then different partial discharge measurement methods will be shown. The advantages, disadvantages and typical fields of application of these methods will also be elaborated upon.
Numerous procedures are used to distinguish between the different partial discharge mechanisms. The most fundamental approach is pulse phase analysis. The cavity discharges happen at zero crossings, surface discharges after zero crossings and the corona discharges around voltage peaks.
Finally I present an algorithm based on a neural network which can categorize the different partial discharge types automatically.