Further Experience in the Use of Existing RDs in Windings of Motors and Generators for the Measurement of Partial Discharges Claude Kane Eaton Electrical Predicative Diagnostics 5421 Feltl Road Suite 190 Minnetonka, MN 55343 claudekane@eaton.com Alexander Golubev Eaton Electrical Predicative Diagnostics 5421 Feltl Road Suite 190 Minnetonka, MN 55343 alexandergolubev@eaton.com Igor Blokhintsev Eaton Electrical Predicative Diagnostics 5421 Feltl Road Suite 190 Minnetonka, MN 55343 igorblokhintsev@eaton.com Abstract: Over six years has passed since the use of the RDs embedded in motor and generator windings as a sensor to detect partial discharges (PD) was introduced. Over this period of time, thousands of measurements utilizing RDs have been made. his coupled with the refinement of measurement techniques and sensor technologies has validated this approach. Use of the existing embedded RDs in conjunction with traditional sensors, such as coupling capacitors at the line terminals, will greatly improve the diagnostics capabilities of insulation systems in rotating equipment. Methodologies and numerous examples will be presented while addressing the strengths and limitations of this technology. INRODUCION Online measurement of partial discharges (PD) has proved to be an effective tool in evaluating the condition of stator insulation in medium voltage electric motors and generators [1]. his method is widely used in addition to the traditional off-line insulation tests performed during scheduled outages. When a PD event occurs a high frequency pulse is induced into the conductor. he frequency spectrum of this pulse ranges from a few kilohertz to a few gigahertz. Most online PD technologies available on the market today measure these PD events in the higher frequency bands, generally one megahertz and higher. One common problem all technologies face is that these high frequency signals attenuate quickly as they travel through the winding. herefore, sensors commonly installed at winding terminals have a limited zone of sensitivity and provide valuable information for that zone only [4]. A generally accepted rule of thumb is that coupling capacitors installed at the line terminal of a machine will only have a zone of sensitivity of 10-15% of the total winding. his equates to the first two or three coils on large motors and hydro generators and maybe the fist bar on large turbine generators. A common sense solution is to use sensors embedded into the windings. Some of the PD technology vendors suggest installing specially designed sensors into the winding, but this approach is relatively expensive and requires an extensive machine outage and invasion into the winding assembly. Alternatively, most of the medium voltage machines already have Restive emperature Device (RD) detectors embedded into the winding by the manufacturer and these detectors can be used for partial discharge measurements [2,4]. Eaton Electrical has over six years of experience using RDs as PD detectors. A special PD transducer was designed that is installed in series with the existing RD wiring at the RD terminal block located on the frame of the motor or generator. he transducer does not affect the operation of the RD or the temperature monitoring equipment connected to the RD. he transducer only passes the high frequency PD signals to the PD instrument. Using RDs as a PD sensor is very effective in trending of machine PD activity when used with an analyzer that can effectively reject noise and process PD data. If sensor calibration is performed, the use of RDs can be further applied to allow comparisons between different machines. WHY RDS WORK A typical motor or generator will have anywhere of 6 to 12 RDs. Many larger turbine generators will have as many as 50 to 60 RDs. Figure 1 shows a typical RD. he RD and the RD wiring will lie in a slot between two coils. In each slot there is a top coil and a bottom coil and may be of the same or different phase. At higher voltages (normally at 6.6 kv and higher) the coils are taped with a semi-conductive tape that keeps RD at ground potential. Figure 1 - Picture of RD found in a typical motor. RD and wiring act as an antenna to capture PD signals occurring in the winding that is not normally detected by traditional line sensors.
How RDs Work As A PD Sensor he RD and the RD wiring act as an antenna. As the PD pulse travels in the vicinity of the RD and the RD wiring the PD energy is coupled to the RD. he pulse then follows the RD wiring out to the PD transducer in common mode (between the RD wires and ground) where the PD pulses are extracted from the wiring and feed to the monitor or analyzer. As the wiring from the RD leaves the stator slot it bundles with the other RD wires in the machine and eventually are brought out to a terminal block on the outside frame of the machine. his is the location where the transducer is installed. Since the RD wiring is intertwined from all the RDs, many of the PD signals will cross couple between all wiring before the signals reach the transducer. In order to make sense out of the PD signals an appropriate analyzer to decipher the signals for analysis is required. Also there is a considerable amount of noise in the RD circuits coming from the RD power supply. he transducer must also reject this noise. Figure 2 shows phase-resolved patterns of the noise coming from the RD power supply before and after installation of the RD PD-transducer. he horizontal axis of each chart is the magnitude of the pulses in millivolts and the vertical axis is the phase angle of the power frequency AC waveform (0-360 degrees). he dots represent the number of pulses per cycle for a given magnitude and phase angle in color scale from black to bright yellow. Figure 2a shows data taken directly from the RDs without the transducer. his noise is coming from the RD power supply. Figure 2b shows the same machine with the RD PD transducer installed. a. b. Figure 2 - Phase resolved data presentation of noise from the RD power supply and PD signals from the machine. a - noise from power supply without transducer b - Noise rejected after transducer is installed and very low PD activity from RD1 is seen (25 mv). One can observe classical PD patterns in Figure 2b, RD1 at 25 millivolts scale after noise from RD power supply is canceled. he transducer is passive in nature and requires no external power and the type and ohm rating of the RD has no affect on the functionality of the system. Figure 3 shows the RD transducer installed on a motor. In this case, the RD terminal block was replaced with the RD transducer. Figure 3 - Installation of the RD transducer in series with the existing RD wiring. Sensitivity of the RD to PD signals can vary based on RD design and wiring methods. his is well known and accepted. A large database of data exists, coupled with calibration tests, which allow for these differences and determination of default values based on these factors. It must be noted that basing a decision on magnitude alone, without calibration can be precarious, no matter what sensor is used. Whether it is a coupling capacitor, a radio frequency current transformer (RFC) or RD. Many improper installations of traditional coupling capacitors have been noted over the past 5 years. Improper installation of the coupling capacitor will affect the magnitude output detected by the analyzer connected to the capacitor. Without calibration, comparison of magnitudes between different machines is frequently very questionable. In the absence of calibration, trending of data, analysis of the phase resolved data (fingerprint or pattern recognition) and behavior of PD activity with humidity, winding temperature, load current, hydrogen pressure, etc, are the key diagnostic and prognostic tools. Even IEEE P1434 [1] realizes that lack of calibration performed by most vendors and stresses trending as the key diagnostic component. Unlike most vendors, Eaton Electrical attempts to perform calibration of every installation, if possible, so absolute magnitude comparisons can be made. Validation of the Use of RDs AS A PD SENSOR Case 1 he effects of signal attenuation discussed above may cause mistakes in evaluating stator winding insulation condition, if sensors located at machine line terminals were the only ones
used for assessment. Figure 4 shows test results of three 13.8kV motors of a similar design at the same facility. All motors have permanent Radio Frequency Current ransformers (RFC) sensors placed on surge capacitor grounding conductors. Figure 4 (top) shows maximum PD magnitudes recorded from the RFC sensors. Based on these results, one can conclude that motor 1 is in good condition and motors 2 and 3 have moderate level of discharges. Data from three RDs for each motor are shown in Figure 4 (bottom). Conclusions, based on this data, are the same for the motors 2 and 3, but the conclusion is different for motor 1. Motor 1 has a high level of PD (nearly 1.3 volts) within the zone of sensitivity of RD01. herefore, this customer now knows where to focus their limited maintenance budget due to having additional information. Motor1 Motor2 Motor3 Motor1 Motor2 Motor3 Figure 4. Comparison of maximum PD magnitudes of RFC s and three RD s. Case 2 RFC Sensors RD Sensors Figure 5 shows data obtained from a 16,000 HP, 13.8 kv that will be further discussed in Case 4. Clamp-on RFCs are installed in the main terminal box on cable shield grounds and 12 RDs and evenly distributed around the winding. On each oscillogram, trace 1 is the signal from the RFC located on phase A, trace 2 and 3 from RD01 and RD03 located on phase A and trace 4 - RD05 corresponding to the middle phase. In Figure 5a, the oscilloscope was triggered from a PD pulse that originates near the line terminals (RFC - Phase A). races 2, 3 and 4 show no response. In Figures 5 b and c the oscilloscope was triggered by a PD pulse originating near RD01 and RD03. AS can be seen, there is no response from the other sensors. his illustrates that PD event originating near the line terminals will be detected by the line terminal sensors, but not by the RDs. Also the RDs and not the line sensors will detect PD occurring near the RDs. herefore, for a complete analysis, it is necessary to install additional sensors into a winding or to use RDs to get information about the winding condition. It is strongly recommended that both traditional coupling capacitors and RDs be used together as they compliment each other very well. 1 > 2 > 4 > 3 > 1) SHA: 100 mvolt 500 ns 2) RD1: 100 mvolt 500 ns 3) RD3: 100 mvolt 500 ns 4) RD5: 100 mvolt 500 ns riggering from line terminal PD. 1 > 2 > 3 > 1 > 2 > 3 > 4 > 4 > 1) SHA: 1 Volt 500 ns 2) RD1: 1 Volt 500 ns 3) RD3: 1 Volt 500 ns 4) RD5: 1 Volt 500 ns 1) SHA: 5 Volt 500 ns 2) RD1: 5 Volt 500 ns 3) RD3: 5 Volt 500 ns 4) RD5: 5 Volt 500 ns riggering from RD01. riggering from RD03. Figure 5 - PD pulse attenuation in a winding of a 16,000 HP, 13.8 kv motor. Case 3 A he data presented in this case is from a 12,000 HP, 13.8 kv machine. Figure 6 shows the before and after reconditioning data from sensors located at the line terminals. he bar graphs show the PD intensity (proportional to the amount of energy in the discharges, calculated in milli-watts), Maximum Magnitude in millivolts and Pulse count in pulses per cycle. Phase resolved data is also shown for these sensors. It is evident that the reconditioning process was effective. Figure 6 - PD intensity, Magnitude, Pulse Count and phase resolved data from line terminals of a 12,000 HP, 13.8 kv machine, before and after winding reconditioning. Figure 7 shows the same data for the same machine from two of the RDs. he reconditioning process was equally effective as illustrated by the data. C B
wo additional points require emphasis. Common misinformation is in the marketplace that all that is detected by the RDs is spurious noise. o substantiate that this is not the case, one needs to just view the phase resolved data in Figure 7 and future case studies presented in this paper. If all that was detected were noise, then there would be no "classical" PD distributions. Classical PD distributions show PD occurring between 0 0-90 0 and 180 0-270 0. 1. If all that was detected were noise, similar patterns would be on all channels. Noise patterns are not sensor specific. 2. If all that was detected were noise, one would see the same patterns both on the before and after data. In the case of Figure 7, it is the same machine, same installation. If there were noise before reconditioning, there would be noise after reconditioning. of around 200-250 mv, which are below normal alarm levels. Figure8 - Phase resolved data from the three RFCs on the termination drain shields of a 16,000 HP, 13.8 kv machine. Figure 9 shows phase-resolved data from six of the 12 RDs. RD01 and RD03 show some extremely high magnitudes approaching 16 volts. All RD data was taken utilizing temporary RD sensors that does not have the ability to filter out the noise from the RD power supply. Figure 7 - PD intensity, Magnitude, Pulse Count and phase resolved data from two RDs of a 12,000 HP, 13.8 kv machine, before and after winding reconditioning. Case 4 his case shows data from a 16,000 HP, 13.8 kv machine from a large petrochemical plant. he client had an upcoming outage and required an assessment as to the condition of the machine prior to the outage. o complicate matters, no traditional PD sensors were installed and an outage could not be scheduled to do so. Split core Radio frequency current transformers (RFCs) were temporally installed around the drain shields of the incoming cable terminations in the motor termination enclosure. he motor also had 12 RDs embedded in the winding. Figure 8 shows the phase resolved data from the three RFCs located at the line terminals. All phases show PD magnitudes Noise from RD Power Supply Figure 9 - Phase resolved data from six of the twelve RDs of a 16,000 HP, 13.8 kv motor without RD ransducer installed. Figure 10 shows the same data as Figure 9, but with the RD transducer installed. Several online tests over a three-month period were performed on this machine as well as off-line tests in a motor repair facility. During the online test, it was noticed that there was a positive correlation between winding temperature and PD activity. his combined with the predominance of energy occurring in the negative half cycle indicated a prognosis that slot exit discharges existed in this machine. Defects located at slot exit points have a high positive correlation with temperature. Upon inspection of the machine in the motor repair facility multiple sites as shown in Figure 11 were evident.
and again in November 2003. ypical PD behavior for a new machine was noticed with a decrease in PD activity over the year time period. Phase resolved data from two of the RDs are shown in Figure 12. Figure 10 - RD phase resolved data with RD transducer installed to eliminate the effects of the noise generated by the RD power supply. Figure 12 - Phase resolved data from two RDs on a 1,200 HP, 4.16 kv machine. Data shows decrease of PD activity over a period of one year. Another example of a 4,160 volt motor is shown in Figure 13. his machine was taken out of service and severe looseness in the wedges was observed. Once again, classical D patterns are observed. Figure 11 - Multiple slot exit discharges were evident upon inspection of the 16,000HP, 13.8 kv machine (damaged coil in the middle). he other interesting aspect was that all defects were not on the high voltage coils of each phase group. here were 6 coils per phase group and in every case, the defects shown in Figure 11 were found on the second and third coil of each phase group. his indicates the RDs were sensitive to the PD signals while the relatively low readings were found on the RFCs located in the line termination enclosure. Also, the RFCs do have a better frequency response than traditional coupling capacitors, therefore, they should see deeper into the winding. Another interesting aspect was that while in the motor repair facility, offline PD tests were performed with the machine at room temperature. Normal operating temperature of the machine while in service was around 110 0 C. here was a significant decrease in PD activity to around 2 volts in magnitude. Use of both a RF meter and an ultrasonic probe showed that all the defects shown in Figure 11 were not active. Case 5 his case shows data from a 1,200 HP 4,160 volts motor located on an offshore platform. he motor was installed in 2002 and data was taken from the RDs in November of 2002 Figure 13 - Phase resolved data from RDs on a 1,300 HP, 4.16 kv machine. SUMMARY his paper addresses the misinformation in the marketplace about the use of RDs as a PD sensor. he value of using RDs as a PD sensor is equivalent to the value of using traditional PD sensors at he line terminals. he paper strongly recommends the use of both types of sensors in order to provide reliable information as to the wellness of the stator insulation of motors and generators. Five case studies were presented validating the data and the use of RDs as a PD sensor. Concerns over sensitivity, noise, data validity and applications were addressed. REFERENCES 1. IEEE P1434 Guide to Measurement of Partial Discharges in Rotating Machinery, 1998. 2. C Kane, A. Golubev, I. Blokhintsev "Advances in the Continuous Monitoring of Partial Discharges in Rotating Equipment", Waterpower November2003, Buffalo, NY 3. C Kane,, J. Pozonsky, S. Carney, I. Blokhintsev " Advantages of Continuous Monitoring of Partial Discharges in Rotating Equipment and Switchgear ", 2003 AISE Meeting, Pittsburgh, PA, September 2003
4. I. Blokhintsev, M. Golovkov, A. Golubev, C. Kane Field Experiences on the Measurement of Partial Discharges on Rotating Equipment, IEEE PES 98, February 1-5, ampa, FL 5. Z. Berler, I. Blokhintsev, A. Golubev, G. Paoletti, A. Romashkov RD as a Valuable ool in Partial Discharge Measurements on Rotating Machines Proceedings of 67th Annual International Conference of Doble Clients, March 27-31, 2000, Watertown, MA 6. Z. Berler, A. Golubev, A. Romashkov, I. Blokhintsev A New Method of Partial Discharge Measurements, CEIDP-98 Conference, Atlanta, GA, October 25-28, 1998.