CURRENT STATE OF SURGE TESTING INDUCTION MACHINES ABSTRACT

CURRENT STATE OF SURGE TESTING INDUCTION MACHINES John Wilson Baker Instrument Company

ABSTRACT Surge testing of motor coils has been an industry practice since J. L. Rylander published “A High Frequency Voltage Test for Insulation of Rotating Electrical Apparatus” in Transactions of the AIEE, in February of 1926. Since that time the use of surge testing to diagnose failing or damaged insulation in motors has advanced greatly. In 1926, the indication of a turn-turn insulation failure was the drop of the coil voltage amplitude as determined by a vacuum tube rectifier circuit using an apparatus the size of a large work bench with rotating spark gaps for switches and large step up transformers to charge large high voltage capacitors. Today’s instrumentation uses high-speed digitizers to observe the entire waveform. Instead of a work bench, a 50lb portable enclosure contains the entire instrument which includes small high voltage power supplies, power semiconductors for switching elements, small efficient high voltage capacitors and a complete Pentium class computer with color display. The principles of surge testing, the strength of the turn-turn insulation being tested, the motivation to perform the surge test, the required test voltages necessary to succeed with the test, the industry standards surrounding the technology and the modern diagnostic algorithms will be presented.

PRINCIPLE OF THE SURGE TEST If a rapidly increasing current is applied to a coil, a voltage will be generated across the coil by the principle of induction. The voltage across the coil is given by V=L*di/dt Where V is the terminal voltage across the coil, L is the coil’s inductance, and The time rate of change of current pulse is di/dt. The terminal voltage V at the leads of the coil is actually a summation of the induced voltage created between individual loops in the coil. If the insulation separating adjacent coils is weak and if the induced voltage is higher than the dielectric strength of the weak insulation, an arc will form between the coils. Surge testing equipment is designed to create the induced voltage between adjacent coils and detect the arcing indicative of weak or failing insulation. Figure 1 below shows a block diagram typical of today’s instrumentation. The internal capacitor is charged to a known voltage by the power supply. At a specific time, a high voltage switch closes which transfers the charge from the capacitor through the windings of the coil. If the resistances and loss of the entire circuit are such that the system is under damped, charge will be able to flow through the inductor and on to the other side of the capacitor resulting in an oscillation This process of ringing will repeat until the resistances and losses in the circuit completely absorb all of the energy that was originally on the capacitor. The measurement of the terminal voltage of the coil vs time gives the surge waveform, which shows the damped oscillation.

Iris Rotating Machine Conference June 2003, Santa Monica, CA

1 of 13

FIGURE 1. Block Diagram of Typical Surge Instrumentation

The ringing frequency of the dampened sinusoidal waveform will be according to the following formula:

f =

1 2π

 1 R2  − 2  LC 4 L

  

If the turn-turn insulation fails with an arcing short between two turns in the coil, a fraction of the inductance will be shorted out of the circuit. Fro m the equation above, the ringing frequency f will increase as the inductance decreases due to the short. An increase in the ringing frequency will show itself to be a jump to the left of the ringing pattern. To reiterate, it is this sudden increase in ringing frequency that is the indication of the arcing turn-turn fault. Depending on the coil and the location of an arcing short, the magnitude of the surge waveform may also slightly decrease. Today’s instrumentation will slowly increase the test voltage and “look” for the increase in ringing frequency.

TURN-TURN INSULATION STRENGTH Dielectric Strength of Round Wire Consider the illustration below showing two round, film covered wires. The film is made of a number of different materials with varying abilities to withstand temperature, with different capabilities to withstand voltage and with different abilities to resist attack by solvents or chemicals.

FIGURE 2. Round Wire Insulation The minimum dielectric strength as required by NEMA MW 1000 is 5700V for a twisted pair of wires. However, wire manufacturers exceed this value by a significant amount. A datasheet for wire typically used in motors will show a dielectric strength of 8,700V at its rated temperature of 200°C. The 8,700V dielectric strength is a high number, especially since this wire is most likely to go into a 480V motor.


2 of 13

Dielectric Strength of Rectangular Wire Rectangular wire is most often used in higher voltage machines (2300V and above) and higher power machines (500hp and above). A three turn coil is illustrated in Figure 3 below and shows the major components that make up the turn-turn insulation.

FIGURE 3. Rectangular Wire Insulation

The rectangular wire has basically the same film insulation material as used in round wire, but often includes a “glass wrap” around the outside of the film. The purpose of this glass layer is to provide a porous material to physically separate two adjacent wires from each other in such a way that resin can fill the space separating the wires. A form coil using rectangular wire is constructed by coiling the wire in a loop with the required number of turns, mechanically forming the coil into the proper shape for the motor and then wrapping the outside of the shaped coil with a tape made of Dacron, mica, or other material. Finally, the coil is placed in the stator, electrical connections are made, and the whole stator and coil assembly is then vacuum impregnated with the resin. The purpose of the tape is to provide an insulation barrier between the rectangular wire and the steel slot of the stator. The tape is made to be permeable to resin so the resin can get through the tape to the underlying wire and impregnate the glass wrap, if present. Resin also fills the porous spaces in the tape increasing its dielectric strength and also mechanically fixes the coil in the stator slot. To determine the turn-turn insulation capability of a form wound coil some assumptions will be made. 1. Rectangular wire with a glass wrap will be used in the calculation. 2. It will be assumed that the resin completely fills the space in the glass wrap. 3. The glass wrap is about 0.005” thick on each wire. 4. The film is the same as round wire insulation from above. 5. The resin is a polyester base and has a dielectric strength of 4600V per 0.001” 6. All materials are physically in contact with each other with no extra space between turns. Using these assumptions, the turn-turn dielectric strength of a form wound coil is: Iris Rotating Machine Conference June 2003, Santa Monica, CA

3 of 13

8700V (film) + [4600V/0.001” * 0.005” (glass wrap filled with resin)] * 2 (two wires)= 54,700V. The assumptions made above can be considered “academic”. The point to be understood is that the turn-turn insulation in well made motors is really quite high. Indeed Gupta et al [1] state that motors can be manufactured with 10pu or higher turn-turn insulation strength (10pu = 38kV on a 4160V motor). Although we have plans for manufacturing coils with 40kV+ turn insulation, nature has different plans for us. Due to inconsistencies in the manufacturing process, the resin does not completely impregnate or fill the glass, the tape, and other void spaces. The result is the appearance of voids in the insulation. These void spaces are where PD can become a major mechanism of premature aging of the insulation. Additionally, winding equipment can damage the wire film or glass wrap exposing bare copper even before the VPI process.

MOTIVATION TO SURGE TEST A basic question to be answered is “Why surge test?” The most direct answer is the only way to detect turnturn insulation weakness is with the surge test. For those who manufacture or rewind motors, the ability to manufacture coils free of insulation defects is paramount to reliability. Therefore, the surge test is used universally in the manufacture of motor coils for small and large motors alike. For motors in service, the dielectric strength of the turn insulation slowly decreases with time as the insulation ages. Some factors causing the insulation to age include: thermal cycling, vibration, mechanical movement of coils abrading insulation, chemical attack, partial discharge, exposure to damaging transients, exposure to radiation, VFD operation, etc. For operators of electric motors, the knowledge that a motor’s insulation is weakened is paramount to maintaining a productive and profitable process. Another related question is: “How good does turn-turn insulation have to be?” In the mid 80’s, Gupta et al [2] monitored 33 motors in 11 power plants over a period of 3 years. Their conclusion was that motors see ~3pu surges in normal operation. Some motors may see surges as high as 4.6pu. (1pu = per unit = sqrt(2/3) * VLine = maximum voltage with respect to ground in a three phase system.) Observed rise times were as low as 0.16µs to over 1µs. The conclusion to be drawn is that a motor must have a minimum turn-turn insulation dielectric strength in order to survive in its environment. If a motor is operated with a severely weakened dielectric strength, a failure of the turnturn insulation quickly results in catastrophic failure of the motor and failure of the process the motor is turning. Regardless of the standpoint of the individual, the knowledge that a motor’s turn-turn insulation is sound is crucial. Therefore, the necessity to surge test is clear.

IEEE 522 RISE TIMES AND TEST VOLTAGES Christiansen et al [3] observed that the turn-turn voltage distribution of a surge pulse is not linear across the windings of a coil being tested. A fast rise time pulse will induce a higher turn-turn voltage at the tested end of the coil than at the grounded end of the coil. The greater turn-turn voltage at the tested end of the coil is due to the wave propagation affects of the high frequency components in the surge impulse as the pulse moves through the winding. Due to this phenomena, IEEE 522 recommends different test voltages for different rise time pulses. For pulse rise times of 0 to 100ns, the recommended test voltage is 1pu. For pulse rise times between 100ns and 1.2µs, the recommended test voltage is 3.5pu. For pulse rise times of 1.2µs or longer, the recommended test voltage is 5pu. The high voltage switching networks in Baker Instrument Company surge testers fall in the 100ns to 1.2µs category. The actual rise times of Baker instrumentation are closer to 100ns. Figure 4 below shows the recommended test voltage given the operating voltage of the motor and the rise time of the surge pulse.


4 of 13

FIGURE 4. Recommended Surge Test Voltage vs Pulse Rise time. (IEEE 522-1992 IEEE Guide for Testing Turn-Turn Insulation on Form-Wound Stator Coils for Alternating-Current Rotating Electric Machines”, Figure 1, page 8.)

RECOMMENDED TEST VOLTAGE The ideal situation for the surge test is to find the arcing turn-turn insulation failure. In such a case the motor insulation is shown to be weak by failing the surge test, yet has not become a welded/hard/bolted short. It can be assumed there is still life left to the insulation, although very little is left. In determining a proper test voltage it must be understood that there is a minimum voltage that must be applied across a gap before there is any chance of an arc forming in a gaseous dielectric. This minimum voltage is called the “Paschen’s minimum”. Figure 5, called Paschen’s Curve, shows the breakdown voltage vs. air spacing at standard temperature and pressure for a uniform field. The minimum voltage is approximately 350V.


5 of 13

FIGURE 5. Paschen’s Curve

Therefore, using the information above, at least 350 volts per turn must be applied to the windings before there is any chance of finding an arcing fault. In reality, the electric field between coils in a motor is not uniform because of field enhancement near sharp corners and variations in insulation. Never the less, it is often recommended that a test voltage apply 500 volts per turn at a minimum. Obviously, there is an upper limit to the voltage used to test an assembled motor. The recommended test voltage of 500V per turn works fine for single coils, but gives a high test voltage for an assembled motor which may have 20-30 turns per phase. One philosophy is to limit the test voltage to a value that applies a significant fraction of the known turn-turn dielectric strength of the coil, yet safely be below the ground wall insulation strength. As seen from the turn-turn dielectric strength calculated above, the turn-turn dielectric strength can be in the 40kV or greater range.

INDUSTRY ACCEPTED SURGE TEST VOLTAGES Several industrial organizations have written standards regarding the application of the surge test. Table 1 below shows the recommended test voltages from the different organizations.

VLine 480 575 600 2300 4160 6900 13800

Per Unit 392 469 490 1878 3397 5634 11268

IEEE 522 New Coils 3.5pu 1372 1643 1715 6573 11888 19718 39437

In Service 75% New 1029 1232 1286 4930 8916 14789 29578

EASA In Service 2*Vline +1000 1960 2150 2200 5600 9320 14800 28600

IEC34-15 1.2 x 50µs 4*Vline +5000 6920 7300 7400 14200 21640 32600 60200

0.2 µs 65% 4498 4745 4810 9230 14066 21190 39130

TABLE 1 Industry Test Voltages


6 of 13

Note there is much variation between the American standards and the European standard. From the discussion above, it is clear to see why. Coils actually can withstand a high, applied turn-turn voltage yet many standards bodies have elected to promote lower test voltages.

VOLTAGE DISTRIBUTION CONCEPTS AND MEASUREMENTS The turn-turn voltage distribution of a form wound coil in air will be very near linear. For example, if a 5000V pulse is applied to a 10 turn coil, then 500V will appear across each turn in the coil. The same coil installed in the steel core of a motor will not have a linear distribution. There are many reasons for this. A simple explanation will be given for what is a very complicated process. Three contributing factors are the speed with which the pulse travels through the winding, the tendency of the iron to attenuate higher frequencies, and the filtering effects of distributed or stray capacitances.

PULSE SPEED The group velocity of energy flowing through conductor mounted in a high permeability iron core is approximately Vg=(3.00 x 10^8)/sqrt(µr * er ) (m/s) Where er is the relative dielectric permitivity of the coil’s electrical insulation and µr is the relative magnetic permeability of the motor’s iron. The µr term is in the 100’s depending on the metallurgy of the steel in the core. The dielectric constant er will be in the range of 2 to 4 depending on the types of materials used in the insulation. In air, µr =1, er~=2 which means the pulse moves through the turns of a coil at very near the speed of light. The pulse is only slightly slowed by the effects of the motor’s turn insulation and reaches each turn in the coil rather quickly – not much after the surge pulse was impressed on the first turn of the coil. Therefore, there won’t be much of a distribution other than that caused by induction (“L*di/dt”). In this case the coil can be considered a lumped element and the normal rules of circuit analysis apply. The speed of the surge pulse in the same coil mounted in the motor’s iron is much slower. The time taken for the surge pulse to enter the first coil, slog its way around the turn and reach the end of the first coil can be on the order of 100ns. Note that 100ns is on the same order of magnitude of the rise time of the applied surge pulse. Therefore, a high turn-turn voltage would be expected in this case because the voltage at the second turn would take some time to appear with respect to the voltage applied at the first turn. The time dependent nature of the turn-turn voltage is in addition to the voltage across the first coil caused by induction (“L * di/dt”). In this situation, the time dependent nature of energy needs to be considered and no longer can the motor coil be considered a lumped element.

IRON ATTENUATION The above explanation still does not fully explain the nonlinear voltage distribution observed in coils. The missing part is that not only does the steel slow down the surge pulse, it also absorbs some of the energy in the pulse and filters the high frequency components. Core steel is lossy at the frequencies of interest in surge testing – 100kHz to 10MHz. For example, for 2% silica steel at 2kHz, the specific loss is in the range of 10-100 watts/lb depending on the applied magnetic field strength. The loss at higher frequencies is even higher. Although this discussion of loss is imprecise, it is clear that loss in the steel is high at the frequencies used in surge testing. The high frequency components in the leading edge of the surge pulse are attenuated more than the loss at the lower frequencies.


7 of 13

FILTERING EFFECTS In addition to the material properties of the steel, there is the filtering effect of the large coil itself. The coil has distributed parasitic capacitances to the core. These parasitic capacitances provide a lower impedance path to ground for the high frequency components of the leading edge of pulse than compared with the lower frequency components that make up the bulk of the surge impulse. Normally, such capacitances are not considered in electrical analysis, but given the large physical size of stators, these factors need to be considered. The data shown below illustrate these effects nicely. A distribution measurement was made on a 4160V coil consisting of three groups of 16 turns per phase of a motor as shown in Figure 6 below

FIGURE 6. 4160V Stator Coil Winding Configuration. Each of the 16 end turns was tapped so that a high speed, high voltage probe could measure the waveform of the surge pulse. The measurements are shown in Figure 7 below where the voltage measurements of first 16 turns plotted simultaneously on top of each other. The overall waveform pattern is that of the lumped element inductance of the machine and the surge tester’s capacitance and is dampened by internal resistances of the surge tester, by the copper coil resistance, and by the loss of the steel.

FIGURE 7. Voltage Distribution Measurement on 4160V Stator Coil


8 of 13

Looking at the leading edge of the surge pulse in detail, the figure shows a separation of waveforms for the first 5 or 6 turns for the first ~20µs of the pulse. From 20µs on, the waveforms for these same coils are neatly nested. It is during the leading edge of the pulse that the effects of lossy iron, propagation delay and filtering all influence the voltage distribution. The remaining turns have their waveforms stacked nicely on top of each other at both the beginning of the pulse and for the remaining duration of the pulse. Figure 8 shows the first 10µs of the differential coil voltage for the same coil shown in Figure 7. The voltage between the 15th coil and the 14th coil is labeled “(C3-15)-(C3-14)”. The coil voltage between the 14th and 13th coil is labeled similarly “(C3-13)-(C3-12)”, like wise with the voltage between the 13th through the 10th coils. The maximum instantaneous voltage is the highest for the first turn “(C3-15)-(C3-14)” with the remaining turns showing a decreasing turn-turn voltage. Note that the maximum inter-turn voltage of 1100V is at the beginning of the pulse where all of the effects described above come into play.

FIGURE 8. Differential Voltage Measurement on 4160V Stator Coil

AUTOMATIC FAULT DETECTION METHODS The greatest advancement in surge testing has come about with use of high-speed analog and digital electronics and the application of computers to control the testing process. Algorithms programmed into the computer can detect small variations in the shape of the waveform that people’s eyes miss. The detection algorithms include: Zero Crossing Shift Pulse to Pulse EAR Line to Line EAR Each of these techniques will be discussed below. An additional benefit to having a computer control the test is the immediate shutdown of the test after the insulation “fails” a single pulse.

ZERO CROSSING SHIFT As described earlier, detecting an arcing turn-turn insulation failure is the ideal situation from the standpoint of maintaining a working motor and its process. To find the arcing turn-turn short, many impulses are applied to a motor each with a slightly higher test voltage. The test voltage begins at “zero” volts and increases to the maximum


9 of 13

test voltage determined by industry standards mentioned above. An arcing turn insulation failure is identified by observing a shift to the left of the surge waveform resulting from a reduction in the inductance of the coil and/or a change in the amplitude or shape of the surge waveform. A change in the zero crossing location of the waveform plot is one indication of an arcing insulation failure.

FIGURE 9. Example of Zero Crossing Failure Figure 9 above shows an example of an arcing turn-turn fault manifested by a zero crossing jump or “Nested Failure”. The two waveforms shown in the figure are of the failed wave and of the previous good wave (also called the previous to fail wave) that is at a slightly lower voltage. The jump in the zero crossing location is pointed out automatically by the zero crossing detection algorithm with the two arrows labeled “Nested Failure”. Unfortunately, the zero crossing algorithm can falsely identify failures due to digitizer noise which makes a person’s eye a much better identifier of the “jump to the left” than a computer.

PULSE TO PULSE EAR Where the computer has the advantage over the human is detecting slight changes of the waveform shape. The computer uses the Error Area Ratio (EAR) to get a quantitative measure of the difference in shape of two subsequent waveforms called the Pulse to Pulse EAR (ppEAR). Before describing the ppEAR, the EAR calculation will be introduced. The formula for the Error Area Ratio [4] is

Npts

EAR1−2 =

∑ Abs(F

i

i =1

− Fi ( 2) )

Npts

∑ Abs (F j =1


(1)

(1 )) j

)

10 of 13

Where Fi (1) is a point in the time series of the first or reference waveform, Fj (2) is the corresponding point in the time series of the second or test waveform, and Npts is the number of points in the time series sampled at each “jth ” point, and EAR1-2 is the Error Area Ratio of the test Fj (2) waveform with respect to the reference Fi (1) waveform. This simple formula is a fast calculation for a computer, yet it is very accurate at detecting a difference in the shape of the two waveforms. Two exactly identical waveforms will have an EAR of 0%. Two waveforms that look identical to the eye can have EAR values in the 4-5% range. Two waveforms with a noticeable difference will have an EAR value in the range of 10% or higher. The application of this formula for the ppEAR is to compare the most recently acquired waveform to the previously acquired waveform as the test voltage is slowly increased. The two waveforms are expected to be different since one is at a slightly higher test voltage than the other. This difference is on the order of 4%. However, if the most recently acquired waveform is above the arcing voltage for the failed insulation, its shape will change. The change in shape which will then be detected by the EAR algorithm. The figure below shows the ppEAR detecting an arcing turn-turn insulation failure.

FIGURE 10. Example of Coil Failing Pulse to Pulse EAR

Figure 10 above shows two surge waveforms and a small ppEAR graph in the lower right corner of the display. The two waveforms represent the previous to fail and the failed waveform. In this case the zero crossing detection algorithm would be hard pressed to find a jump in the waveform, because the jump to the left is so small. The small graph shows the running ppEAR where the first value plotted is for the two pulses at 800V and 810V. The ppEAR values for successive pulses, each 10 volts higher than the previous pulse, are shown along the majority of the graph until a pulse is applied above the dielectric strength (~2700V) for the weak insulation. The ppEAR for the failed pulse compared to the previous to fail pulse is shown completely off the scale on the graph. The dashed line along the top of the graph represents the user defined maximum allowable ppEAR for the test. Anything above this value will fail the test. In the figure above, the maximum allowable ppEAR was set to 5% and the “failed” ppEAR was 12%. The ability to perform such detection schemes as the ppEAR is a testament to the advancement of the surge testing instrumentation available today. The ability to precisely control the impulse voltage, to precisely trigger the Iris Rotating Machine Conference June 2003, Santa Monica, CA

11 of 13

impulse generator, and the ability to implement advanced signal processing algorithms during the testing proces s are key to being able to find the very small changes in surge waveforms that represent insulation failures. An area of research is detection of partial discharge events on the surge waveform using the ppEAR technique to detect the change in shape of the waveform due to PD events.

LINE TO LINE EAR The Line to Line Error Area Ratio (L L EAR) is a way to quantitatively determine the difference of the waveforms from two phases of a motor. The thought is that the surge waveforms for all three phases of a motor should be nearly identical. If the waveforms are not, there must be a hard or bolted short in one of the phases. The LL EAR works well for motors without the rotors installed because the coils are nearly identical. Testing assembled motors (rotor installed) is not so straight forward. The problem with the LL EAR is in the assumption that the surge waveforms are all identical for assembled motors. This is not the case due to unequal rotor coupling between the phases. Recall the surge test is applied off line with the rotor at a stand still. Depending on where the rotor slots are with respect to the slots of the stator coil being tested, different coupling ratios will be observed between the three different phases. Therefore, a different surge waveform will be observed on all three phases. Some motors show this problem more so than others. Some motors don’t have the problem at all. Unfortunately, there is no way to tell beforehand if a motor will couple un-equally. Also, there is the issue of lap and concentric wound machines, which have slightly different coils in each phase. The LL EAR detection algorithm is often left active when testing assembled motors so the computer running the test has another “failure criteria” to determine insulation soundness. However, the failure criteria is raised to a level above the point where unequal rotor coupling would fail the winding.

PARTIAL DISCHARGE DETECTION ON SURGE WAVEFORMS Iris Power Engineering has developed an instrument, the PDAlert, that can identify the existence of partial discharge on a surge pulse. The first of the two figures below (Figures 11 and 12) show the output of the instrument testing a motor that does not exhibit PD. The second figure shows the typical PD signature obtained with this instrument.

FIGURE 11. PD Free Motor Iris Rotating Machine Conference June 2003, Santa Monica, CA

12 of 13

FIGURE 12. Motor with PD

Note the difference between the instrument’s output at the beginning of the pulse and where the PD event occurs. The signal of the PD is clearly separable from the leading edge transients caused by the fast rise time surge pulse. It is remarkable that the small amplitude PD signature can be identified in the presence of the large amplitude surge pulse, even though the bandwidths of the PD and surge pulse are similar.

CONCLUSION Surge testing has come a long way from the day the technology was invented. Today’s instrumentation is able to find deeper faults in windings, is lighter and more portable, and is automated both in operation and in fault detection. As a result manufacturers of motors can make a better, more reliable product and the users of motors to operate their processes at a higher up time and with a higher degree of certainty.

REFERENCES [1] Gupta, B.K., Lloyd , B.A., Stone, G.C., Sharma, K.K, Fitzgerald, J.P., “Turn Insulation Capability of Large AC Motors, Part 2 – Impulse Strength”, IEEE Transactions on Energy Conversion, Vol. EC-2, No. 4, pp 666-673, December 1987. [2] Gupta, B.K., Lloyd , B.A., Stone, G.C., Campbell, S.R., Sharma, K.K, Nilsson , N.E., “Turn Insulation Capability of Large AC Motors, Part 1 – Surge Monitoring”, IEEE Transactions on Energy Conversion, Vol. EC-2, No. 4, pp. 674-679, December 1987. [3] Christiansen, K.A., Pedersen, A., “An Experimental Study of Impulse Voltage Phenomena in a Large AC Motor”, IEEE Publication 68C6-EI-87, pp. 148-150, 1968. [4] United States Patent Number 5,111,149, “Method and Apparatus for Automatically Calculating the Integrity of an Electric Coil”. Iris Rotating Machine Conference June 2003, Santa Monica, CA

13 of 13

CURRENT STATE OF SURGE TESTING INDUCTION MACHINES ABSTRACT

Recommend Documents