270 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 27, NO. 2, JUNE 2012 Detection of Artificial Insulation Defects in a Medium-Voltage Motor by Dielectric Spectroscopy Analysis E. Obame, P. Rain, O. Gallot-Lavall´ ee, and G. Tripot Abstract—Artificial defects have been introduced during the manufacturing process inside the composite insulation of stator models of medium-voltage motors dedicated to the application of traction. The dielectric analysis at low frequency (10 −3 –10 3 Hz) has been conducted at 30 ◦ C for each of the seven different models. The sensitivity of the dielectric properties to humidity is illustrated. This is the reason why the measurements were made with the mod- els in a dried state. The dielectric analysis allowed the detection of most of the seven defects sought. These defects have also been iden- tified. The value of the capacitance at 1 kHz appeared in particular as a reliable marker for identification. The observed differences be- tween models with defect and the reference model are qualitatively justified by dielectric considerations. Index Terms—AC motors, composite insulation, dielectric mea- surements, insulation testing. I. INTRODUCTION T HE insulation of a medium-voltage rotating machine is a complex multilayered impregnated system. The qual- ity of the final product requires following a given number of steps during the manufacturing process. Sound control insula- tion materials during the manufacturing process are necessary to prevent premature aging. Many studies have reported on the detection and identification of defects resulting from the aging of the insulation. However, very few published works have been devoted so far to possible defects in new insulations. To the extent that manufacturing imperfections lead to insulation defects similar to those that may result from aging, they should be detected and identified by the same methods. Many techniques have been developed to assess the quality of the insulation in the stator windings of large power machines Manuscript received May 31, 2011; revised October 6, 2011; accepted December 21, 2011. Date of publication January 24, 2012; date of current version May 18, 2012. Paper no. TEC-00263-2011. E. Obame was with Grenoble Electrical Engineering Laboratory (G2Elab), Grenoble University, CNRS, 38042 Grenoble Cedex 9, France. He is now with the High Voltage Engineering Group, Department of Electric Power En- gineering, Chalmers University of Technology, SE 412 96 Gothenburg Sweden (e-mail: elysee@chalmers.se). P. Rain and O. Gallot-Lavall´ ee are with the Grenoble Electrical Engi- neering Laboratory (G2Elab), Grenoble University, CNRS, 38042 Grenoble Cedex 9, France (e-mail: pascal.rain@grenoble.cnrs.fr; olivier.gallot-lavallee@ grenoble.cnrs.fr). G. Tripot is with ALSTOM Transport, 25290 Ornans, France (e-mail: gerard. tripot@alstom.transport.com). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TEC.2011.2182052 [1]–[3]. Most of them have also been applied to medium-voltage motors, as was explained in detail in [4] and [5]. These methods are generally carried out to detect possible deterioration of the equipment and to decide on a conditional maintenance. Under specific conditions in terms of voltage and dimensions, voids resulting from aging cause partial discharges, which may contribute to the degradation of the insulation. Partial-discharge measurement techniques have proven relevant to detect and identify these kinds of defects and have been applied to mo- tors of a few kilovolts [6]–[8]. The low-frequency voltages are especially relevant to detect anomalous conduction phenomena. The use of a constant dc voltage, of a dc ramp, or of a low-frequency ac voltage has been successful to detect insulation defects responsible for modifica- tions in the conduction properties, such as moisture, contami- nants, or a lack of curing [9]–[11]. Various indicators are used to support the diagnosis and take the inferred decisions. The most commonly used are the insulation resistance (IR) after 1 min, the polarization index (PI), and the comparison of polarization and depolarization currents (PDC). Since the defects considered in this study mainly consist of voids and variations in dielectric constants, the impedance spec- troscopy appears to be a promising technique. In addition, di- electric spectroscopy can operate at low voltage without the risk of harm to the material. The aim of this study is to investigate whether dielectric spectroscopy allows the detection of specific artificial defects. Seven types of defects have been selected and seven models of stators have been built, each with one defect. The dielectric spectra of each defective model will be compared to the spectra of the reference model in order to consider the ability of defect detection. Qualitative analysis of the observed deviations will be proposed for each defect. Eventually, all the spectra will be compared for identification purpose. Due to the influence of humidity on the dielectric properties, which is detailed in Section IIIA, the measurements have been conducted on models in a dried state. II. EXPERIMENTAL DETAILS A. Models 1) Materials: The insulation was designed for stator coils of ac traction motors. The model consists of a section of a stator with an inversed curvature as shown in Fig. 1. There are six slots with two bars in each slot (see Fig. 2). Each bar is composed of five copper flat wires covered with a polyimide Kapton (see 0885-8969/$31.00 © 2012 IEEE