Detection of Faults in Rotor-Windings of Turbogenerators

Conference on Diagnostics in Electrical Engineering CDEE 2016

Detection of Faults in Rotor-Windings of Turbogenerators Dr. Christian Staubach, Stefan Krane Siemens AG, Muelheim an der Ruhr, Germany Abstract— This paper should give an overview of different measurement methods for detecting faults in rotorwindings of turbogenerators and explains their principal functionality in detail. Advantages and disadvantages of each individual measurement are summarized and compared. In addition theoretical calculation models to evaluate the specific methods are illustrated while at the same time the accuracy of this theoretical models are validated with the help of practical testing. Keywords—rotor winding; short detection; turbogenerator diagnosis; RSO; flux probe; c-core; voltage pole drop

I.

INTRODUCTION

For full and safe operation rotor-windings of turbogenerators have to be insulated towards ground and between each neighboring turns. During manufacturing, revisions and operation several methods for faultdetecting exist to validate the functionality of the insulation. Some of these methods not only detect faults but also allow for positioning these faults within the winding. A ground fault, i.e. a short between the winding and the (grounded) rotor body, can be easily detected with an insulation resistance measurement or a high voltage test. Much harder to detect is a short (mainly ohmic transition contacts) between two neighboring turns or two neighboring coils. II.

CAUSES AND EFFECTS OF SHORTS

Possible causes of faults in rotor windings are ageing of the insulation due to relative movement of the coils within the slot (figure 1a), contamination that cause an ohmic contact between individual turns (figure 1b) or loosening of blocking within the endwinding area that may lead to plastic deformation of the windings and the coils (figure 1c).

depending on the conductivity of the short and compensated by an increased current, which will result in an asymmetrical temperature distribution. The magnetic field distortion will result in asymmetrical forces and increased vibration level of rotor of a turbogenerator. III.

TEST METHODS

This paper only covers standardized tests that are performed during manufacturing, revisions and operations to detect faults in windings. The test methods can be modified to position an already detected fault within the winding that particularly work well if individual coils and turns can be contacted e.g. with a measurement tip. In principle, methods can be divided into online diagnosis (performed during operation) and offline diagnosis (performed during standstill). For online diagnosis the evaluation of magnetic flux within the air gap between rotor and stator is commonly used. The flux is evaluated with the help of an induced voltage in a small coil (“flux probe sensor”). Reliable offline test methods include the winding resistance measurement, the impedance measurement (at power or elevated frequency), the voltage surge measurement (“Recurrent Surge Oscillography”), the voltage drop measurement and the “C-Core”-measurement. Impedance and pole balance tests that depend on rotational frequency can only be performed under specific conditions, e.g. when a spin-pit is available. IV.

ONLINE DIAGNOSIS

A. Flux probe measurement An air-core coil can be used as a method to detect a winding fault within the rotor in the so called flux-probe measurement [2]. This method only works if the rotor is excited, e.g. during operation. Due to the turning of the excited rotor during operation a sinusoidal alternating voltage is applied to the air-core coil that is permanently attached to a stator tooth as shown in figure 2.

Fig. 1. Electrical faults in rotor windings: (a) damaged insulation, (b) contaminated winding, (c) deformed endwinding-coils [1]

During operation these shorts between turns in the rotor winding may reduce the amplitude of the generated magnetic field as well as generate an asymmetry in field distribution. The quantity of this decrease is strongly 978-1-5090-6179-2/16/$31.00 ©2016 IEEE

Fig. 2. Installation of an air-gap flux probe to the stator tooth [3-4]

When analyzing the signal from the air-core coil shortened turns in the rotor winding can be detected when superposition the signals for each pole winding due to their


symmetrical structure. For evaluation the measurement signal is divided into sections of defined time periods that can be used for further calculations. Figure 3 shows a calculation of the total current through the slot/nut and radial flux signal when no short is present and figure 4 shows results when a short is present.

are applied separately to the voltage test prods. The principle setup of the test method is shown in figure 5.

Fig. 5. (a) Test setup of R20 measurement, (b) test device [1]

A shortened turn causes a lower R20 value. The measurement can be executed easy and fast with highly portable and space-saving equipment. Another advantage is that the R20 is also an important factor to determine the quality of cold soldering joints and to detect corrosion of conductors, cracks and material abrasion of conductors. This also can lead to misinterpretations since the measurement is very sensitive of transition resistances. With this test method only dead shorts can be detected and positioning of the fault is not possible. Fig. 3. Calculated measurement signal of flux probe when no short is present [1]

B. Impedance measurement For the impedance measurement AC voltage (with power or elevated frequency) is applied to the terminals of the winding and the current through the complete winding is measured as shown in figure 6a. The impedance should not change over operation time if there are no faults in the winding and should be comparable to measurements in the same condition in the past. If there is a short present the impedance should be reduced compared to a prior measurement (without a shortened turn). In the spin-pit a shortened winding that is only present during turn operation can be detected by measuring the impedance while accelerating the rotor up to nominal speed. If a speed dependent short is present a hysteresis impedance curve can be observed as shown in principle in figure 6b.

Fig. 4. Calculated measurement signal of flux probe when short is present in coil C of pole 1 [1]

The advantage of the flux probe measurement is that it can be performed during operation and allows positioning of the fault within the winding. The sensitivity of the measurements depends on the load (MVA and cosφ) and determining a boundary value is very complex and depending on the fault location. V.

OFFLINE DIAGNOSIS

A. Winding resistance measurement R20 The winding resistance is measured at the terminals of the winding with direct current. The measurement results are temperature dependent and therefore have to be corrected to 20°C to result in the R20. Since it is important to not include the resistance of the transition from the current supply clamps of the measurement device, the winding resistance measurement is carried out using 4 wire resistance measurement, where the current supply clamps

Fig. 6. (a) Test setup of impedance measurement, (b) typical impedance curve if speed dependent fault is present [1]

Advantages for this measurement are higher sensitivity due to possible elevated frequency and proximate modelling of electrical operation conditions. Disadvantages include the necessity of high apparent power as well as complex limit values or acceptance criteria due to dependability of design and possible fault. C. Voltage surge measurement Another one of the methods to identify shorts in the rotor winding system is the “Recurrent Surge Oscillography (RSO)”. A voltage peak is applied from both sides of the rotor winding connections consecutively and the response is noted. Figure 7a shows the principle


test setup and figure 7b a winding test device that can be used to conduct the measurement.

Fig. 7. (a) Test setup of voltage drop measurement, (b) test device [1]

In principle the measurement can be carried out using a high voltage pulse of 1000-2500 V as used in the RSO and also a low voltage pulse of 5-10 V. A real test setup for the RSO measurement is shown in figure 8 with the connections for the terminals of the rotor and the test unit.

Fig. 10. Response signals of RSO with short in coil B of pole 2 [5]

Caused by damping effects and dispersion of deviating fault positions with different electrical properties the evaluation of the RSO response signals is very complex, requires a lot of experience and is empirically not distinctly possible. Nevertheless it is possible to determine the double pass transit time (DPTT) of the signal through the winding as shown in figure 11 and use the findings to position a fault within the rotor winding.

Fig. 8. (a) Connection for appliing surge to winding, (b) test unit [1]

Theoretically due to the symmetrical structure of the rotor winding and the corresponding signal transit times these response signals should overlap perfectly if no fault is present. Figure 9 shows an example measurement for a winding where no fault is present. Fig. 11. Signals for determining the double pass transit time (DPTT)

With the RSO measurement method it is also possible to detect non galvanic faults like displaced winding insulation.

Fig. 9. Response signals of RSO when no fault is present [5]

Figure 10 respectively shows response signals where an artificial short is applied within the rotor winding that lead to a separation of the two recorded signal responses from each pole at a specific time. This can be evaluated and allows conclusions to position the fault within the winding of the rotor.

D. Voltage pole drop measurement The voltage pole drop measurement relies on the symmetrical setup of the rotor winding. A low-frequency sinusoidal voltage signal with defined amplitude is applied to the winding terminals and the voltage drop across each individual pole winding is measured respectively between terminal and mid-point of the winding, the so called pole crossing. The principle test setup is shown in figure 12.

Fig. 12. Test setup of voltage pole drop measurement [1]


In theory the voltage drop for a two-pole rotor should be equal between both pole winding terminals to the midpoint when no fault is present. A deviation in the voltage drop values can be caused by a shortened turn, i.e. a pole with a shortened turn should have a lower voltage drop compared to the other pole. Figure 13 shows pole drop measurement results of a rotor without fault at standstill and under rotation as well as results for two measurements with artificial shorts in the winding under rotation.

VI.

TABLE I. Frequency Range 1) 0 Hz

R20 Impedance

Fig. 13. Voltage pole drop measurement at standstill and under rotation in spin pit without a short and with artificial shorts

Due to already mentioned influences the interpretation of deviations can cause incorrect conclusions. The voltage drop measurement can be conducted with net frequency or elevated frequency similar to the impedance measurement. E. C-Core measurement For the “C-Core” measurement at DC current is applied to the rotor winding and the voltage at the c-core device monitored while turning. A principle test setup is shown in figure 14.

Fig. 14. Test setup for c-core measurement [1]

CONCLUSION

The presented test methods show that there is not a single preferred method for detecting faults in windings of turbogenerator-rotors. The measurements in summary or a combination of test methods can lead to a thoroughly diagnosis of a functioning rotor. The following table 1 gives a summary overview of the individual test methods including an estimation of necessary apparent power that is often connected to the desired mobility of test equipment.

50…500 Hz

TEST METHODS OVERVIEW AND C OMPARISON Detect Position

Under Rotation

no

no

no

approx. Power 10 – 100 W

Limit value

Online Monitoring

yes

no

yes

0.1 – 10 kVA

yes

yes

0.1 – 10 kVA

yes

no

Pole Balance

50…500 Hz

no

yes 2)

RSO

Pulse

yes

yes

-

no

yes 3)

FluxProbe

0 Hz

yes

yes

-

yes

yes

yes

no

no

no

C-Core

0 Hz

yes

yes 2)

10 – 100 W

Pole Drop

50…500 Hz

yes

no

0.1 – 10 kVA 1)

50 Hz region 2) spin pit 3) theoretically

Operators and manufacturers often use different test methods with diverse parameters so that measurements are comparable not at all or to a limited extend. Even when using the same test setup and parameters some measurements show dependencies on winding structure and fault location that make fixed limited values impossible to define. It is rather necessary to analyze a measurement in detail to interpret deviations in measurement signals or values and draw the right conclusions. This can also be the reason why in international standards few of these methods can be found. Nevertheless it is possible to standardize the tests with an adequate freedom of parameters to provide operators and manufacturers with common guidelines and an accepted theoretical basis to allow maximum comparability and individual determination of acceptance criteria or limit values. By means of latest and powerful numeric field calculation the opportunity arises to establish databases for different types of windings, faults and fault locations that aid a fast analysis of conspicuous or abnormal measurement results.

If a shortened turn is present in one nut the integrated voltage trend will be lower than in the other nuts.

REFERENCES [1]

[2] [3] Fig. 15. Integrated voltage trend for c-coil measurement when short is present in nut 2 [1]

[4]

Even though this measurement is only possible in the spin pit under rotation it has the benefit that it is possible to determine an acceptance criteria and limit value.

[5]

C. Staubach, R. Merte; Windungsschlussanalyse an Rotoren von Turbogeneratoren, 7. Essener Tagung: Turbogeneratoren in Kraftwerken, Essen, Germany, 2014. D. R. Albright; Interturn short-circuit detector for turbinegenerator rotor winding, 1970. Generatortech Inc.; Generator Field Winding Shorted Turn Detection Technology, 2006. M. Sasic, B. Lloyd, A. Elez; Finite Element Analysis of Turbine Generator Rotor Winding Shorted Turns, IEEE Transactions on Energy Conversion, 2012. S. Krane; Reflektionsmessung als Diagnoseverfahren zur Detektion und Ortung von Wicklungsfehlern bei Läufern von Turbogeneratoren, Masterarbeit, WH Gelsenkirchen, 2015