INTRODUCTION
LARGE generators play a key role in power systems. Because they are expensive, unexpected breakdowns represent huge economic losses. High quality maintenance is therefore essential to ensure a high level of availability. In short, it is very important to ensure that large generators are in good repair and can run stably The vibration of stator bars caused by electromagnetic force exists in the operation of large generators and this is worsened by loose stator slot wedges. The severe vibration accelerates the deterioration of the ground wall insulation and ultimately causes generators to fail, so it is important to detect the state of a generator’s stator slot wedges in order to ensure the generator is working safely. The main method of assessing slot wedge tightness is to tap the slot wedge by hand using a small hammer, with the operator making a judgment of tightness or looseness based on the sound. This method is workable and has provided much useful data about the condition of the generators, but has a few of major drawbacks. First, the method is very subjective, as it depends on the individual operator’s judgment of the sound coming from the slot wedge. Also, the operators have to work in cramped conditions for long periods of time. Finally,all of the data is handwritten and has to be either studied in its original form or manually entered into a computer for analysis

[attachment=8583]

EARLIER METHODS

In recent years, some companies or groups have researched methods for checking slot wedges for large generators. The First Hydro Company, which is part of Edison Mission Energy and Industrial Development Bangor (UCNW) Ltd., North Wales, made a semi-automatic slot wedge tapping gun in 1995. Its operation was similar to manual slot wedge tapping, with the operator holding and operating the gun by hand. The device was large and heavy. It was also hard to use; It required a great deal of preparation work and a device had to be installed in the generators for slot wedge assessment.
Westinghouse Electric Corporation introduced a slot wedge assessment device that can acquire displacement data of a slot wedge after an impact of a certain force on it. After an impact of known force has been imposed on a slot wedge, a curve of force displacement is formed and compared to the standard curves acquired from standard lab slot wedge tests to determine whether the slot wedge is tight or loose. That device is not very exactness for the vibration amplitude of slot wedges is very little at the beginning loose stage. This paper is concerned with specific loosening process of slot wedges with different fixing styles, and presents judgments of a slot wedge’s tightness or looseness based on its vibration characteristics during the loosening process. Moreover, an automated detection system was developed with a digital signal processor (DSP) of Texas Instrument Company (TI), based on the analysis of loosening process of slot wedges, and a series of tests were performed in our lab with it to detect the looseness of slot wedge samples of a stator slot model. The results soundly verified the correctness of the judgments and the validity of the detection system .


ANALYSIS OF LOOSENING PROCESS



Large generators can be sorted into two classes generally; in one type, the stator slot wedges are fixed with battens, and in the other, they are fixed with ripple springs. Under the same impact, the vibration characteristics of slot wedges fixed with battens are very different from slot wedges fixed with ripple springs.


SLOT WEDGE FIXED WITH BATTEN

Figure 1 is a sketch of a stator slot showing the conductor bars, slot wedge and the batten located under the slot wedge. The slot wedge and batten are manufactured from epoxy glass and the batten provides the wedging pressure in the slot.


Figure 1. Sketch of stator slot wedge fixed with batten


When generators become operational, their slot wedges are tightly fixed against conductor bars in the slot. However, slot wedges will become looser and looser as time goes by because generators vibrate. The loosening process of slot wedges fixed with battens could be described as a sequence of four stages, which are shown in Figure 2. The shaded area in the figure indicates that the slot wedge is still tightly fixed at this part, while the white area represents the loose part of the slot wedge




Loosening process of stator slot wedge fixed with batten.

When generators become operational, their slot wedges are tightly fixed against conductor bars in the slot. However, slot wedges will become looser and looser as time goes by because generators vibrate. The loosening process of slot wedges fixed with battens could be described as a sequence of four stages ,which are shown in Figure 2. The shaded area in the figure indicates that the slot wedge is still tightly fixed at this part, while the white area represents the loose part of the slot wedge. The details of the loosening process are as follows:

A. A loose part first appears under the slot wedge after along period of operation.

B. The loose part extends further gradually under electromagnetic force.

C. A loose space appears between the slot wedge and stator bar when the loose part extends so much that it eventually is completely under the whole slot wedge.

D. The loose space becomes deeper and deeper, and finally the slot wedge even falls
off

In stage A, not only can the whole body of the slot wedge and stator bar begin to vibrate under an impact, but also the loose part can begin to vibrate as well. When the impact is a constant force, the frequency of the loose part will diminish gradually as the loose part extends further (Stage B) while the frequency of the whole body would remain the same since there is no change with the entity.
The displacement of the loose part is very small at the beginning stages of slot wedge loosening. But in stages C and D, since the slot wedge has been totally separated from the stator bar as a result of the appearance of a loose space between them, the vibration frequency of the slot wedge remains constant but its displacement increases evidently as the loose space becomes deeper








Figure 2. Loosening process of stator slot wedge fixed with batten.




A loose stator slot wedge was simplified to a thin board model based on elastohydrodynamic theory to analyze the vibration characteristics theoretically. As shown in Figure 3,the coordinate x-y plane coincides with the middle plane of the slot wedge, and the dimension of its loose part is set to a on the x-axis and b on the y-axis. Since rigorous development of the free vibration equation of the board is too involved to be given here, the resulting equation will merely be written as:


where w is the displacement of the board, q is its weight on the area unit, g is the acceleration of gravity and D is the flexural rigidity of the board. The inherent frequency of the board could be given by solving equation (1) as [13]

From the expression above, it could be figured out that the frequency of the loose part will diminish as the loose part extends further.

Figure 3. Thin plate model of a loose slot wedge fixed with batten

1: slot wedge
2: stator bar
3: tight part
4: loose part.


SLOT WEDGE FIXED WITH RIPPLE SPRING








Figure 4. Sketch of stator slot wedge fixed with ripple spring.


Figure 4 shows a sketch of a stator slot showing the conductor bars, slot wedge and the ripple spring located under the slot wedge. The slot wedge and ripple spring are also manufactured from epoxy glass and the compression of the ripple spring provides the wedging pressure in the slot .The loosening of slot wedges fixed with ripple springs is much simpler than that of slot wedges fixed with battens described above, and is illustrated in Figure 5. When generators become operational, the slot wedges are fixed tightly with a great amount of compression in the ripple springs, providing huge wedging pressure. The compression gradually decreases as a result of vibration in the operation of generators.




Because of the ripple spring under the slot wedge ,there is always a bounce force engendered by it against the slot wedge whenever how loose the slot wedge is. The value of the bounce force decreases as the slot wedge becomes looser and, as a result of this, the vibration frequency of the slot wedge decreases and its displacement increases accordingly when an impact of constant force is performed on the slot wedge.



(a) slot wedge fixed tightly (b) slot wedge fixed loose
Figure 5. Loosening process of stator slot wedge fixed with ripple spring.

STATOR SLOT WEDGE TEST SYSTEM

The test system was developed independently in our lab based on the loosening process analysis described above. It is a portable test system, and easier for workers to use to judge the tightness of slot wedges than the traditional manual method. The test system mainly consists of microphone and displacement sensors, an electronic tapper, a DSP micro system and a display and control interface. Figure 6 illustrates how the test system is used to test slot wedges.


Figure 6. Schematic diagram of stator slot wedge test system.

The electronic tapper is used to tap the slot wedge with a constant force (set by the operator) with a very short pulse to cause it to vibrate. This step is carried out according to instructions given by the display and control interface. Next ,the sensors simultaneously send data about the vibration to the DSP micro system to process




Two graphs are generated from the data; one is a displacement graph from the displacement sensor and the other is a frequency spectrum from the Fast Fourier Transform (FFT) of the time-domain data received from the microphone. Finally, the DSP micro system judges the slot wedge’s looseness based on the data of the two graphs intelligently and, displays the results of the test on the display and control interface. Figure 7 is a flowchart showing how the test system works.




Figure 7. Flowchart of stator slot wedge test system.



MAKEUP OF THE STATOR SLOT MODEL

The test specimens were stator slot wedges, battens and ripple springs of an 18 kV/300 MW generator manufactured by the Beijing Steam Turbine Generator Co., Ltd. The stator bar specimen was cut from an entire stator bar from the generator, and the slot model was made of steel to simulate the stator slot of the generator. Since it was much heavier than the stator bar, it remained stationary when a short pulse was applied to the slot wedge, while the slot wedge and the bar would begin to vibrate. Figure 8 shows the details of the specimens and the setup of the stator slot model



MEASUREMENT AND DISCUSSION


TESTS OF SLOT WEDGE FIXED WITH BATTEN


To verify the vibration characteristics in Stages A and B ,six samples with the loose part at a length of 0 mm, 50 mm,80 mm, 120 mm, 160 mm, 200 mm, respectively and at a thickness of 7 mm each were placed along the slot. The six frequency spectrums acquired from the test system are shown in Figure 9. It is clear that there are two frequencies in each spectrum except the first one (the sample with a loose part length of 0mm, which means that it is fixed tightly).
One is the vibration frequency of the whole body of the slot wedge and stator bar (called the main frequency) and the other is the frequency of the loose part (called the partial frequency).Table 2 shows the values of the main frequency and partial frequency for every sample. When the two sets of frequencies from the six spectrums were plotted on a graph with the loose part length as the x-axis, it can be identified that value of the main frequency remained nearly the same, while the value of the partial frequency linearly decreased when the length of the loose part became longer (see Figure 10).Figure 11 shows the vibration characteristics of two samples with loose parts at the same length of 80mm and thicknesses of 4 mm and 7 mm respectively.
It is clear from Table 3 that the amplitude (0.012 and 0.023, respectively) at partial frequency increases as the thickness of the loose part gets longer, but the value of the partial frequency remains nearly constant (756.9 Hz and 770.85 Hz, respectively), as does as the value of the main frequency (1044.51 Hz and1051.62 Hz, respectively).Therefore, in Stage A and B, based on the frequency spectrum, the loose part of the slot wedge could be identified according to the length by the value of the partial frequency and according to the thickness by the amplitude at the partial frequency




FREQUENCY SPECTRUM OF SLOT WEDGES WITH VARIOUS LOOSE PART LENGTH




MAIN FREQUENCY AND PARTIAL FREQUENCY






Frequency spectrum of slot wedges with various space thickness

FREQUENCY SPECTRUM AND DISPLACEMENT GRAPH OF SLOT WEDGES AT DIFFERENT STAGES

In Stages C and D, there were two samples with apparent different thicknesses of loose space under measurement. The curves in Figure 12 show that there was only one frequency(the inherent frequency of the whole slot wedge) in each frequency spectrum and the value of the frequency remained nearly the same; however, the amplitude of the displacement varied greatly when the status of the slot wedge varies from a little loose to loose. The curves also show that the maximum amplitude of displacement was 0.005 (nearly zero) when theslot wedge was moderately loose, while it was 0.288 when the slot wedge was loose. The displacement graph definitively indicated whether the slot wedge loose or tight.




FREQUENCY SPECTRUM AND DISPLACEMENT GRAPHS OF SLOT WEDGES WITH
RIPPLE SPRING.




TESTS OF SLOT WEDGE FIXED WITH RIPPLE
SPRING



To investigate the vibration characteristics of slot wedges with ripple springs, samples were placed in the slot in three apparently different statuses: loose, moderately loose and tight. Since the slot wedge was fixed with a ripple spring ,there was only one frequency in each spectrum: that is, the frequency of the slot wedge with pressure on it. When the status of the slot wedge changed, the wedging pressure changed along with the amount of compression in the ripple spring consequently and thus, the vibration frequency changed. Figure 13 shows the frequency spectrums and displacement graphs of the three samples with different degree of looseness .Table 4 indicates that the vibration frequency of the slot wedge increases as the slot wedge becomes tighter, but the maximum amplitude of displacement diminishes. Both of them vary obviously, which means that in large generators with slot wedges fixed with ripple springs, loose slot wedge detection can be carried out on the vibration frequency spectrum of the slot wedge or its vibration displacement graph.








CONCLUSION



This paper presented a novel solution for stator slot wedge testing in large generators. First, an analysis of the slot wedge loosening process showed ways looseness can be determined for different wedge types .For slot wedges fixed with a batten, the determination oft he tightness of the slot wedge is mainly based on the frequency spectrum of the slot wedge because its displacement is too slight to distinguish in the early stages(Stages A and B) of the loosening process. However, because the frequency is stable in the later stages, the looseness of the slot wedge can be analyzed in Stages C and D according to the slot wedge’s displacement. For slot wedges fixed with ripple spring, the determination of the tightness of the slot wedge could depend on both the frequency spectrum and the displacement characteristics of the slot wedge. It is easier to determine the tightness of a slot wedge fixed with a ripple spring than a slot wedge fixed with a batten. This study focused on the vibration characteristics of stator slot wedges with different fixing styles during the loosening process, and established several ways to determine the tightness of a slot wedge. Moreover, an automatic detection system was developed to detect loose slot wedges in large generators. The proposed system has several advantages compared with the manual slot wedge check or other previous slot wedge testing techniques.

REFERENCES


[1] Z. Jia, Y. Hao, and H. Xie, “The Degradation Assessment ofEpoxy/Mica Insulation under Multi-stresses Aging”, IEEE Trans .Dielectr. Electr . Insul, Vol. 13, pp. 415-442, 2006.

[2] Y. Hao, G. Wang, and Z. Jia, “Application of ultrasonic pulse-echo method to insulation condition diagnosis for large generators”, IEEETrans. Dielectr. Electr. Insul, Vol. 12, pp. 104-113, 2005.

[3] J.K. Nelson, and S. Azizi- Ghannad, “Measures and technologies to enhance the insulation condition monitoring of large electricalmachines”, IEEE Trans. Dielectr. Electr. Insul, Vol. 11, pp. 102-112,2004.

[4] D.H. Locke, and A. Williams, “Stator slot wedge testing of large pumped storage generator motors”, First IEE/IMech E International Conference on Power Station Maintenance-Profitability Through Reliability, Edinburgh, 1998.

[5] A.J. Brown, “Stator winding assessment using TVA probe”, Discharges in Large Machines (Digest No. 1998/264) , IEE Colloquium, pp. 6/1 -6/3, London, Mar. 1998.

[6] S. Cherukupalli, R.A. Huber and C. Tan, “Application of some novel non-destructive diagnostic tests for condition assessment of stator coil sand bars following voltage endurance tests”, IEEE Intern. Sympos.