Recent Important Changes in IEEE Motor and Generator Winding Insulation Diagnostic Testing Standards

Transcription

1 Recent Important Changes in IEEE Motor and Generator Winding Insulation Diagnostic Testing Standards Copyright Material IEEE Paper No. PCIC-2004-XX Greg C. Stone IEEE Fellow Iris Power Engineering 1 Westside Drive, Unit 2 Toronto, Ontario M9C 1B2 Canada Abstract - IEEE standards and test procedures are widely used by motor and generator vendors and users to commission windings in new machines, as well as evaluate the condition of the winding insulation in operating machines. Until recent revisions, the basic procedures and standards in use were written over 25 years ago. Since the 1970s, motor windings have encountered many changes in their design and manufacture. The result was that the interpretation of results in many of the standards was no longer valid for the more modern motors. Over the past 5 years, the IEEE Power Engineering Society has conducted a major review and updating of most of these standards. Many important changes in test procedures and interpretation guidelines have resulted. This paper reviews the main insulation standards used for stator and rotor winding diagnostic testing, and discusses the changes that have been made. Standards discussed include: IEEE 43, 56, 95, 286, 522, and For example, IEEE now requires a minimum insulation resistance of 100 Megohms for new stator windings rated 2300 V or more, rather than the kv+1 that was required in the past. Furthermore, the interpretation for polarization index has changed such that a motor with a PI of 1 is no longer automatically classed as bad. Index Terms stator windings, insulation, testing, diagnostic tests, DC Hipot, AC Hipot. I. INTRODUCTION For most motors and generators, the expected life of a stator winding depends on the ability of the electrical insulation to prevent winding faults. That is, the need for a stator rewind is almost always determined by when the electrical insulation is no longer able to fulfill its purpose, rather than, for example, being determined by a problem with the copper conductors. This follows from the fact that the electrical insulation has a large organic content, a lower melting temperature, and a lower mechanical strength that the copper and the core steel. Thus when a new stator winding is made, NEMA MG1 and IEC require an extensive array of testing to performed on the electrical insulation, to ensure that the stator winding will achieve a satisfactory service life typically years. In addition, many tests have been standardized to be used to evaluate the condition of the insulation during service. The former tests are called acceptance or commissioning tests. The latter are maintenance tests. Over many decades, the Power Engineering Society of the IEEE has been active in documenting most of the stator winding insulation diagnostic tests now used throughout the world by both machine manufactures and machine users. Today, the IEEE lists the following as active insulation test standards, recommended practices and guides that are used by both manufacturers and users [1-6]: 1) IEEE : insulation resistance and polarization index (new and aged windings) 2) IEEE : AC hipot tests (aged windings) 3) IEEE : DC hipot tests (new and aged windings) 4) IEEE : power factor tip-up tests (new and aged windings) 5) IEEE : hipot tests for turn insulation (new and aged windings) 6) IEEE : partial discharge tests (new and aged windings) Most of the IEEE guides, practices, and standards intended for both manufacturers and users, were originally written in the 1950s, and were significantly updated in the 1970s. There have been few changes to these standards and guides for the past 20 years, in spite of the significant changes in stator winding insulation systems. Some of those changes are the following: 1. The widespread application of Class F epoxy and polyester systems.

2 2. The expansion of the global VPI process such that virtually all motors are now made using complete impregnation of the stator core. Starting in the mid 1990s, and continuing to today, IEEE working groups have been completely revising the IEEE testing documents to reflect the changing nature of motor and generator insulation systems. This paper summarizes the purpose of each test method, provides some theory about the test, discusses pass/fail criteria where appropriate, and outlines differences from past versions of the IEEE standard, practice or guide. Additional information on these and other tests is also presented in Reference [7]. II. IEEE 43 INSULATION RESISTANCE AND POLARIZATION INDEX This is probably the most widely used diagnostic test for motor and generator rotor and stator windings. It can be applied to all machines and windings, with the exception of the squirrel cage induction motor rotor winding, which does not have any insulation to test. This test successfully locates pollution and contamination problems in windings. In older insulation systems, the test can also detect thermal deterioration. Insulation resistance (IR) and polarization index (PI) tests have been in use for more than 70 years. Both tests are performed with the same instrument, and are usually done at the same time. The last revision to IEEE 43 was in A. Purpose and Theory The IR test measures the resistance of the electrical insulation between the copper conductors and the core of the stator or the rotor. Ideally this resistance is infinite, since after all, the purpose of the insulation is to block current flow between the copper and the core. In practice, the IR is not infinitely high. Usually, the lower the insulation resistance, the more likely it is that there is a problem with the insulation. PI is a variation of the IR test. PI is the ratio of the IR measured after voltage has been applied for 10 minutes (R 10 ) to the IR measured after one minute (R 1 ), i.e.: PI = R 10 /R 1 A low PI indicates that a winding may be contaminated with oil, dirt, insects, etc. or soaked with water. In the test, a relatively high DC voltage is applied between the winding copper and the stator or rotor core (usually via the machine frame). The current flowing in the circuit is then measured. The insulation resistance (Rt) at time t is then: Rt = V/It which is just Ohm s law. V is the applied DC voltage from the tester, and I t is the total current measured after t minutes. The reference to the time of current measurement is needed since the current is usually not constant. There are four currents that may flow when a DC voltage is applied to the winding. These four are: 1. Capacitive current. When a DC voltage is applied to a capacitor, a high charging current first flows, then decays exponentially. The size of the capacitor and the internal resistance of the voltage supply, typically a few hundred kilohms, sets the current decay rate. A motor stator winding may have a total capacitance of about 100 nf. Thus this current effectively decays to zero in less than 10 seconds. Since this capacitive current contains little diagnostic information, the initial insulation resistance is measured once the capacitive current has decayed. This time before taking the current reading has been set as one minute to ensure that this current does not distort the insulation resistance calculation. 2. Conduction current. This current is due to electrons or ions that migrate across the insulation bulk, between the copper and the core. This is a galvanic current through the groundwall. Such a current can flow if the groundwall has absorbed moisture, which can happen on the older thermoplastic insulation systems, or if a modern insulation has been soaked in water for many days or weeks. This current also flows if there are cracks, cuts or pinholes in the ground insulation (or magnet wire insulation in random wound machines), and some contamination is present to allow current to flow. This current is constant with time, and ideally is zero. With modern insulation, this current usually is zero (as long as there are no cuts, etc) since electrons and ions cannot penetrate through modern epoxy -mica or film insulation. Older asphaltic mica insulations always had non-zero conduction currents, since such insulation systems absorb moisture. If this current is significant, then the winding insulation has a problem. 3. Surface leakage current. This is a constant DC current that flows over the surface of the insulation. It is caused by partly conductive contamination (oil or moisture mixed with dust, dirt, fly ash, chemicals, etc.) on the surface of the windings. Ideally this leakage current is zero. However, if this current is large, it is likely that contamination induced deterioration (electrical tracking) can occur. This current can be large in round rotor windings where

3 the copper conductors are bare, and the insulation is just slot liners. 4. Absorption current. This current is due to a precessing (re-orientation) of certain types of polar molecules in the applied DC electric field. Many practical insulating materials contain polar molecules that have an internal electric field due to the distribution of electrons within the molecule. For example, water molecules are very polar. When an electric field is applied across water, the water molecules all align, just as magnetic domains become aligned in a magnetic field. The energy required to align the molecules comes from the current in the DC voltage supply. Once the molecules are all aligned, the current stops. This current is the polarization current, which is one component of the absorption current. There are many polar molecules in asphalt, mica, polyester and epoxy. Experience shows that after a DC electric field is applied to such materials, the absorption current is first relatively high, and decays to zero after about 10 minutes. In all practical respects the absorption current behaves like an RC circuit with a long time constant. The absorption current, like the capacitive current, is neither good nor bad. It is merely a property of the insulation materials. In addition to molecular re -alignment, absorption currents may aris e in high voltage laminated insulation (such as in high voltage stator groundwalls), due to electron trapping at interfaces. The total current I t is the sum of all these current components. Unfortunately, each of these component currents cannot be directly measured. The currents that are of interest, as far as a winding condition assessment is concerned, are the leakage and conduction currents. If just R 1 is measured (after 1 minute), the absorption current is still non-zero. However, if the total current is low enough, then R 1 may still be considered satisfactory. Unfortunately, just measuring R 1 has proved to be unreliable, since it is not trendable over time. The reason is that IR is strongly dependant on temperature. A 10 o C increase in temperature can reduce R 1 by 5 to 10 times. Worse, the effect of temperature is different for each insulation material and type of contamination. Although some temperature correction graphs and formulae are in the IEEE 43, they are acknowledged as being unreliable for extrapolation by more then 10C or so [1]. The net result is that every time R 1 is measured at different temperatures, one gets a completely different R 1. This makes it impossible to define a scientifically acceptable R 1 over a wide range of temp eratures. It also makes trending R 1 almost useless, unless one can be sure the measurement temperature is always the same. The polarization index (PI) was developed to make interpretation less sensitive to temperature. PI is a ratio of the IR at two different times. If we assume that R 10 and R 1 were measured with the winding at the same temperature, which is usually very reasonable to assume, then the temperature correction factor will be the same for both R 1 and R 10, and will be ratioed out. Thus PI is relatively insensitive to temperature. Furthermore, PI effectively allows us to use the absorption current as a yard stick to see if the leakage and conduction currents are excessive. If these latter currents are much larger than the absorption current, the ratio will be about one. Experience shows that if the PI is about one, then the leakage and conduction currents are large enough that electrical tracking will occur. Conversely, if the leakage and conduction currents are low compared to the absorption current after 1 minute, then PI will be greater than 2, and experience indicates that electrical tracking problems are unlikely. Thus, if we can see the decay in the total current in the interval between 1 minute and 10 minutes, then this decay mu st be due to the absorption current (since the leakage and conduction currents are constant with time), with the implication that the leakage and conduction currents are minor. B. Test Method The IR is measured with a high voltage DC supply and a sensitive ammeter. The DC supply must have a wellregulated voltage; otherwise a steady state capacitive charging current will flow. The ammeter must measure currents smaller than a nanoamp. There are several special purpose megohmeters available commercially. Sometimes these are known as Megger Testers, after the name of the instrument first developed for this purpose (Megger is a trade name of AVO). A megohmeter incorporates a regulated DC supply and an ammeter that is calibrated in megohms. Modern instruments can apply voltages exceeding 10 kv, and measure resistances higher than 100 GΩ?. The IR and PI test results will depend strongly on the humidity. If the winding temperature is below the dew point, there is no way that R 1 and R 10 or PI can be corrected for the humidity. If the results are poor, then the test must be repeated with the winding above the dew point. It will probably be necessary to heat the winding in some fashion, sometimes for several days, to dry off the moisture that has condensed on the winding. IEEE suggests the IR and PI tests be performed with the winding heated above the dew point. IEEE suggests that test voltages be higher than recommended in the past, because tests at higher voltages are more likely to find major defects such as cuts through the insulation in the endwindings. Note that the test voltages are still well below the rated peak line-to-ground voltages of the

4 windings. Thus the IR test is not a hipot test. Table 1 shows the suggested test voltages. Table 1 Guidelines for dc voltages to be applied during Winding rated voltage (V) * *Rated line-to-line voltage for three-phase ac machines, line-toground voltage for single-phase ac machines, line-to-ground, voltage for single-phase machines, and rated direct voltage for dc machines or field windings. C. Interpretation Insulation resistance test direct voltage (V) < > What constitutes a good reading and a bad reading depends on the nature of the insulation system and the component (stator or rotor) being tested. Until 2000, the minimum R 1 and the acceptable range for PI was essentially the same for all types of stator winding insulation. However, it has been recognized that the modern insulation materials in random wound and form wound stators have essentially no conduction current (as long as there are no cracks or pinholes). Thus it is possible for a clean, dry, form wound stator winding to have an R 1 that is essentially infinite greater than 100 GΩ. With an R 1 of infinity, calculations of a realistic PI are dubious. Such high R 1 s are not likely in systems made before the 1970 s. Consequently, the maintenance person needs to establish the type of insulation used in the winding, or at least the approximate age of the winding, before interpreting IR and PI results. 2. The minimum R 1 is the value corrected to 40 o C. Unfortunately, any more than o C correction is unlikely to be valid. 3. The minimum acceptable R 1 is much lower for old stators than new stators, and it depends on voltage class. For modern stators, the minimum acceptable R 1 only depends on whether it is a form wound or random wound stator. 4. For modern form wound stators, if a very high R 1 is measured (say greater than 5 GΩ), then PI is not likely to indicate anything about the winding. Thus, one can save time by aborting the test after the first minute of testing. 5. If the IR or PI is below the minimum in a modern stator winding, it is only an indication that the winding is contaminated or soaked with water. 6. If a high PI result is obtained on an older stator winding, then there is a possibility the insulation has suffered thermal deterioration. This occurs because thermal deterioration fundamentally changes the nature of the insulation, and thus the absorption currents that flow. The insulation has changed in an asphaltic mica winding if the asphalt has been heated enough to flow out of the groundwall. Table 2-Recommended minimum insulation resistance values at 400C (all values in MW ) Minimum Insulation Resistance R1 min = kv+1 TEST SPECIMEN For most windings made before about 1970, all field windings, and others not described below Table 2 summarizes how to interpret IR and PI results in stator and rotor windings. The distinction between older and modern insulation systems was set at 1970, although this is somewhat arbitrary. Of note in this table: R1 min = 100 For most dc armature and ac windings built after about 1970 (form wound coils) 1. If R 1 is below the indicated minimu m, the implication is that the winding should not be subjected to a hipot test, or be returned to service, since failure may occur. Of course if historical experience indicates that a low R 1 is always obtained on a particular winding, then the machine can probably be returned to service with little risk of failure. R1 min = 5 Notes For most machines with random -wound stator coils and form-wound coils rated below 1kV 1 - IR 1 min is the recommended minimum insulation resistance, in megohms, at 400C entire machine winding 2 - kv is the rated machine terminal to terminal voltage, in rms kv

5 In general, the IR and PI tests are an excellent means of finding windings that are contaminated or soaked with moisture. Of course the tests are also good at detecting major flaws where the insulation is cracked or has been cut through. In form wound stators using thermoplastic insulation systems, the tests can also detect thermal deterioration. Unfortunately, there is no evidence that thermal deterioration or problems such as loose coils in the slot, can be found in modern windings [1, 7]. III. IEEE 56 MAINTENANCE AC HIPOT TEST IEEE 56 is an extensive guide on various tests and inspections that can be performed on rotor and stator windings, as well as a review of the major repair methods. The document saw its last major revision in 1977 [2], and is now the subject of a complete revision by a working group that is combining IEEE 56 with IEEE 432, so that one guide will cover all form wound motors and generators. The revised version of the standard will probably be published in Although IEEE 56 discuses many tests, of relevance here is the maintenance AC hipot. A hipot test is a high potential applied to the winding. In order to find gross flaws in the winding, the high potential test voltage is normally higher than what the winding sees in service. The basic idea is that if the winding does not fail as a result of the high test voltage, the winding is not likely to fail anytime soon due to insulation aging when it is returned to service. If a winding fails the AC hipot test, then a repair or rewind is mandatory, since the groundwall insulation has been punctured. The AC hipot is similar to the DC hipot (section IV), with the exception that power frequency (50 or 60 Hz) voltage is used. Sometimes 0.1 Hz AC is also employed, as described in IEEE 433. Both commissioning (acceptance) and maintenance AC hipot versions of the test are in use. This test is most commonly applied to form wound stator windings. The maintenance AC hipot is rarely used in North America. A. Purpose and Theory The purpose of this test is to determine if there are any major flaws in the groundwall insulation, before a winding enters service (commissioning or acceptance hipot test) or during service (maintenance hipot test). The principle is that if there is a major flaw in the insulation, a high enough voltage applied to the winding will cause insulation breakdown at the flaw. By IEC and NEMA MG1 standards, all new windings (original or rewound) are subjected to a successful hipot test prior to being accepted by the customer. Of course the main problem with hipot testing (both AC and DC see the next Section) is that the winding may fail. If failure does occur, then either: 1. The insulation that punctured must be replaced. 2. The coil with the puncture is removed from the circuit. 3. The coil or even the complete winding is replaced. These are all expensive alternatives, and all involve a delay in placing the machine in service. Since a hipot test can be destructive and delay a return to service, many people decide not to perform a maintenance AC hipot. The rationale is that the hipot test may cause a failure that would not occur for a long time in service, resulting in rewinding or significant repairs before they are really needed. This is true. However, the proponents of hipot testing argue that for many critical machines, an in-service failure (that could have been prevented if a hipot test was done) can result in a greater disruption to plant output than a hipot failure. For example, the in-service failure of a critical pump motor in a petroleum refinery can stop production for days or weeks, and cost as much as $1M per day. Also, an in-service fault can sometimes cause consequential damage such as stator core damage, a fire or coils being ejected from the slot, resulting in much higher repair costs. Thus, whether an AC hipot is performed as a maintenance test depends on how critical the machine is to the plant, the availability of spares, and the philosophy of plant management to avoid unexpected plant shutdowns. With the AC hipot, the voltage distribution across the thickness of the groundwall insulation is the same as the distribution in service since the applied voltage is AC, and capacitances determine the distribution. NEMA MG1 and IEC define the AC acceptance hipot level as 2E + 1 kv, where E is the rated rms phase-tophase voltage of the stator. IEEE 56 recommends the AC maintenance hipot be 1.25 to 1.5E [2], and this is unlikely to change in the current revision. For example, if the guidelines in IEEE 56 are used, the AC hipot test voltage for a 4.1kV motor would be about 6kV rms. The hipot test is applied between the copper conductor and the stator or rotor core. The AC hipot will age the insulation. In most cases, the hipot voltage is sufficiently high that significant partial discharge activity will occur. These partial discharges will tend to degrade the organic components in the groundwall, thus reducing life. However, calculations based on IEEE 930 indicate that insulation deterioration from a 1-minute AC hipot test at 1.5E is equivalent to about 235 hours or 10 days at normal operating voltage. Therefore, the life is not

6 significantly reduced by a hipot test if the expected life is about 30 years. B. Test Method The key element in an AC hipot test is the AC transformer needed to energize the capacitance of the winding. A 13.8 kv motor stator winding with a capacitance C of 1 µf, requires a charging current of 8 A at f = 60 Hz for a V = 1.5E maintenance hipot test (I = 2? fcv). A minimum transformer rating is over 150 kva. This is a substantial transformer, and is definitely not very portable as compared to a DC hipot set. The AC hipot set is also much more expensive than the DC supply. It is because of the size and expense of the AC hipot supply that an AC hipot is rarely performed as a maintenance test in North America. C. Interpretation A winding either passes or fails the AC hipot. There is no other diagnostic information provided. If the winding fails, as determined by the power supply circuit breaker tripping, then repairs, coil or winding replacement is required. A. Purpose and Theory IV. IEEE 95 - DC Hipot Test IEEE describes the test methods and suggests tests voltages for the DC hipot test [3]. There are differences between a DC and an AC hipot test. Most of the description given for the AC hipot test described above is relevant for the DC hipot. Specifically, the DC hipot is a go-no go test that ensures that major insulation flaws which are likely to cause an in-service fault in the near future, can be detected in an offline test. The previous version of IEEE 95 was published in The major difference between the DC and AC test is the test voltage applied, and how the voltage distribution across the groundwall insulation. Both are linked. With DC voltage, the voltage dropped across insulation components within the groundwall and in the endwinding depends on the resistances (resistivity) of the components. Components with a lower resistance will have less voltage dropped across them. In contrast, the AC voltage dropped across each component in the groundwall or in the endwinding depends on the capacitance (dielectric constant) of each component. Thus, there tends to be a completely different electric stress distribution across the groundwall between AC and DC tests. In older insulation systems, particularly asphaltic -mica systems, the differences between the AC and DC stress distributions were less pronounced because of the finite resis tivity in older groundwalls due to the absorption of moisture. However, with modern epoxy mica insulations, the resistivity is essentially infinite, thus the DC voltage may all be dropped across a very thin layer of insulation. Consequently, significant flaws may not result in puncture with a DC test that would be easily detected with an AC test because of the more uniform voltage distribution with AC stress. For modern windings, the AC hipot test yields an electric stress distribution across the groundwall thickness that is the same as occurs during normal operation. Consequently, the AC hipot is more likely to find flaws that could result in an in-service stator failure if a phase-to-ground fault occurs in the power system, causing an over voltage in the unfaulted phases. For this reason, the AC hipot is superior to the DC hipot, especially with modern thermoset insulation systems. In the 1950s there was considerable research on the relationship between AC and DC hipot tests, and specifically the ratio of the DC to AC hipot voltages [3, 7]. Eventually a consensus was reached that, under most conditions, the DC breakdown voltage is about 1.7 times higher than the AC rms breakdown voltage. This relationship has been standardized in IEEE 95. This research was based on older insulation systems, and unfortunately is largely irrelevant in modern insulation systems, since, as described above, the voltage distribution is completely different under AC and DC. There have, however, been a few studies of the relationship between AC and DC breakdown in modern groundwall insulation systems. One of the largest of these studies pointed out that the ratio of DC to AC breakdown voltage on average was 4.3 in epoxy mica insulation [8]. The 1.7 factor, then, no longer seems to be valid, but since the variability is so large, no replacement ratio has been proposed. Thus the 1.7 ratio is maintained in the latest version of IEEE 95. B. Test Methods There are several different methods for performing a DC hipot. Most are reviewed in IEEE Standard 95, and the 2002 version highlights a new variation of the DC hipot called the DC Ramp test. Some of the variations reduce the risk of a failure during the test, and some also give information of a diagnostic nature. For all types of DC maintenance hipot test methods, the critical decision to be made is the maximum test voltage. For form wound stator windings, IEEE 95 gives guidance. It suggests that the maintenance hipot should be 75% of the acceptance hipot level. NEMA MG1 and IEC stipulate that the DC acceptance hipot be 1.7 times the AC hipot acceptance level of 2E+1 kv, where E is the rated rms phase-to-phase voltage of the stator winding. After performing the arithmetic, it works out that the DC maintenance hipot level should be about 2E. That is, a 4.1 kv winding would be tested at about 8 kv, DC. This level was originally suggested since it approximates the highest likely over voltage in the motor that can occur if a phase-toground fault occurs in the powe r system. Consequently, a

7 maintenance hipot just reproduces, in a controlled, off-line fashion, the over-voltage a stator can see in service. The idea here is that if the winding can survive this hipot, it is unlikely to fail in service due to a voltage surge created by a power system fault. The DC hipot does not age the winding insulation since partial discharges occur very infrequently under DC voltage. Thus, if the winding passes the DC hipot, then the insulation has not been deteriorated in any way by the test. However, one should be aware that if the DC hipot test is done from the switchgear, and if the power cables have been soaked in water for years, then the DC hipot might age and even fail the power cables. This occurs because power cables rated 2300 V and above often fail by a mechanism called water treeing. [9]. A DC potential accelerates water treeing. If the cables have always been kept dry, then DC hipot testing should pose no risk to the cables. There are several alternative DC hipot test methods. 1. Conventional DC Hipot In the conventional maintenance DC hipot, a suitable high voltage DC power supply (available from many suppliers) is connected to the winding, either at the switchgear, or at the machine terminals. The DC voltage is quickly raised to the test voltage and held for either 1 minute or 5 minutes. After this time, the voltage is quickly lowered, and the winding is grounded. If the insulation is sound, there will be no high current surge, and the power supply circuit breakers will not trip. If the power supply breaker trips, then it is likely a puncture has occurred, since the insulation resistance will have instantaneously dropped to zero, which causes an infinite current to flow (by Ohm s law), and the power supply can not deliver this infinite current. Circuit breaker tripping is an indication that the winding has failed and winding repairs or replacement is required. The conventional test contains little diagnostic value, although one can measure the DC current after the 1- or 5-minute application of the test voltage. If one trends the leakage current over the years, then an increasing trend is an indication that contamination is occurring. 2. Step-Stress Hipot A variation is to use the same supply as described previously, and gradually increase the voltage in either equal or unequal steps. For example, the DC voltage can be increased in 1 kv steps, with each voltage level being held for 1 minute before it is increased again. One then measures the DC current after the end of each step (since by this time the capacitive current will have dropped to zero), and plots it on a graph of current versus DC voltage. Ideally, the plot will be a line with a gentle upward curve. However, sometimes the current increases abruptly above a certain voltage. This may be a warning that the insulation is close to puncturing. If the tester acts rapidly, the test can be aborted (voltage turned off) before a complete puncture occurs. Experience shows that warning is likely if the flaw is in the endwinding, but little or no warning is given if the flaw is within the slot. By carefully applying this test, a hipot failure may be avoided. However, if the voltage at which the current instability was detected is below operating voltage, there is a high risk in returning the winding to service without repairs. 3. DC Ramp Hipot A third variation of the DC hipot is called the Ramp test. In this case, the DC voltage is smoothly and linearly increased at a constant rate, usually 1 or 2 kv/minute. Thus, there are no discrete steps in voltage or current. The current vs. voltage plot is automatically graphed and displayed. By increasing the voltage as a constant ramp, the capacitive current is a constant current which can be easily ignored, unlike in the stepped stress test. The primary advantage of the ramp test is that it is by far the most sensitive way to detect when a current instability is occurring, since the capacitive charging current is not changing with time. Consequently, the ramp test is the method most likely to enable the user to avoid a puncture, and it may even enable detection of significant delamination [10, 11]. A DC Ramp test unit is now commercially available. C. Interpretation Fundamentally, the DC hipot test is not a diagnostic test that gives a relative indicator of the insulation condition. Rather it is a go-no go test, where the winding is in good condition if it passes, and in severely deteriorated condition if it fails. However, the DC current measured at the time of the test can give some qualitative indication of condition, much like the IR and PI tests. Specifically, if the current at any particular voltage increases continuously over the years, it is an indication that the insulation resistance is decreasing, and the winding is gradually getting wetter or becoming more contaminated. However, caution is needed when trending the current over time. The current is very dependent on winding temperature and atmospheric humidity. Thus, in most cases the trend is erratic, and impossible to interpret. V. IEEE 286 POWER FACTOR TIP-UP IEEE 286 [4] describes the power factor tip-up test, sometimes called the dissipation factor tip-up test. Tip-up is an indirect way of determining if partial discharges (PD) are occurring in a high voltage stator winding. Since PD is a symptom of many high voltage winding insulation deterioration mechanisms [6, 7], the tip-up test can indicate if many failure processes are occurring. In addition to its use as an off-line maintenance test, the tip-up test is widely used by stator coil and bar manufacturers

8 as a quality control test to ensure proper impregnation by epoxy and polyester, during coil manufacture. IEEE did not change interpretation, but did align the IEEE document with the IEC equivalent IEC A. Purpose and Theory All practical insulating materials have a dielectric loss, which can be measured with a power factor (PF) or dissipation factor (DF) test. The DF and PF measurement methods are different, and yield slightly different results. At low voltages, the PF and DF are not dependent on voltage. However, as the AC voltage is increased across the insulation in a form wound coil, and if voids are present within the groundwall, then at some voltage partial discharges will occur. These discharges produce heat, light and sound, all of which consume energy. This energy must be provided from the power supply. Consequently, in a coil with delaminated insulation (perhaps due to long term overheating), as the voltage increases and PD starts to occur, the DF and PF will increase above the normal level due to dielectric loss, since the PD constitutes an additional loss component in the insulation. The greater the increase in PF and DF, the more energy that is being consumed by the partial discharge. In the tip-up test, the PF or DF is measured at a minimum of two voltage levels. The low voltage PF, PF lv, is an indicator of the normal dielectric losses in the insulation. This is usually measured at about 20% of the rated line-to-ground voltage of the stator. The voltage is then raised to the rated line-to-ground voltage, and PF hv is measured. The tip-up is then: tip-up = PFhv - PFlv The higher the tip-up, the greater is the energy consumed by PD. Some organizations will record the PF or DF at several different voltage levels, and calculate several different tip-ups between different levels. By plotting the tip-up as a function of voltage, the voltage at which PD starts is sometimes measurable. If the PF or DF is measured in percent, then the tip-up is in percent. Historically the test was first applied to high voltage stator bars and coils, to ensure that the insulation was completely impregnated. However, since the late 1950s, some motor and generator operators have applied the test to complete windings to detect various aging mechanisms that produce PD. Measurement of the tip-up is complicated by the presence of silicon carbide stress control coatings on coils rated 6 kv or above. At low voltage, the silicon carbide is essentially a very high resistance coating, and no current flows through it. Thus there is no power loss in the coating. However, when tested at rated voltage, by design the silicon carbide coating will have a relativity low resistance. Capacitive charging currents flow through the insulation and then through the coating. The charging currents flowing through the resistance of the coating produce an I 2 R loss in the coating. The DF or PF measuring device measures this loss. Since the loss is zero at low voltage, and non-zero at operating voltage, the coating yields its own contribution to tip -up. It is not uncommon for the tip-up due to the stress relief coating to be 2 or 3%. This coating tip-up creates a minimum tip-up level. Very significant PD must be occurring in most windings for the PD loss to be seen above the silicon carbide tip-up. When manufacturers test individual coils and bars in the factory as a quality assurance test, the tip-up contribution due to the stress relief coating can be negated by guarding [4]. Unfortunately, it is not practical to guard out the coating tip-up in complete windings. See IEEE 286 for specific test methods. B. Interpretation As a maintenance tool for complete windings, the tip-up test is used for trending. The initial value of the tip-up on a phase is of little significance, because it will be dominated by the stress relief coating contribution to tip-up. However, if the tip-up is measured every few years and the tip-up starts increasing from the normal level, then it is likely that the winding has significant PD activity. To increase the tip-up above the normal level requires widespread PD. The most likely causes of this PD are: 1. Thermal deterioration 2. Load cyclin g 3. Improper impregnation during manufacture The tip-up test is not likely to be sensitive to loose coils in the slot, semiconductive coating failure or endwinding electrical tracking. In all these cases the PD is at a relativity low repetition rate, or the damage is confined to the relatively small portion of the winding, and thus the PD contribution to tip-up is relatively minor. VI. IEEE 522 SURGE TESTS None of the tests discussed above directly measure the integrity of the turn insulation in form wound or random wound stator windings. The stator voltage surge test described in IEEE does this by applying a relatively high voltage surge between the turns. This test is a hipot test for the turn insulation, and may fail the insulation, requiring a repair, coil replacement or rewind. The test is valid for any random wound or multi-turn form wound stator. IEEE 522 is currently being revised, but the changes are likely to be mainly be clarifications.

9 A. Purpose and Theory Switching on a motor causes a fast risetime voltage surge to hit the stator winding terminals. Similar voltage surges occur from IFDs and faults in the power system. These fast risetime surges result in a non-uniform voltage distribution across the turns in the stator winding. If the risetime is short enough, the surge voltage high enough, and the turn insulation weak enough, then the turn insulation punctures, rapidly leading to a stator ground fault. The surge test duplicates this action of an external surge. As such, this test is analogous to the AC and DC hipot tests: apply a high voltage to the turn insulation, and see if it fails. The surge test is a destructive go-no go test. If the turn insulation fails, then the assumption is that the stator would fail in service due to motor switch-on, IFD surges or transients caused by power system faults. If the winding does not puncture, then the assumption is that the turn insulation will survive any likely surge occurring in service over the next few years. Thus the main question is whether a maintenance surge test should be performed or not, and this is a philosophical question identical to that posed for the AC and DC hipot. The main difficulty with the surge test is determining when turn insulation puncture has occurred. In the DC or AC hipot test, groundwall puncture results in the insulation resistance plummeting to 0 Ω. This causes the power supply current to increase dramatically, opening the power supply circuit breaker. There is no question that puncture has occurred. A turn-to-turn puncture in a winding does not cause a huge increase in current from the power supply. In fact, if there are 50 turns between the phase terminals and neutral, the failure of one turn will only slightly reduce the inductive imp edance of the winding, since the impedance of only one turn has been eliminated. Thus the other 49 turns can continue to impede current flow, and the circuit breaker does not trip. In the surge test, turn failure is detected by means of the change in res onant frequency caused by shorting out one turn. The inductor is the inductance of one phase of a stator winding, or in the motor stator where the neutral ends can not be isolated, the inductance of (say) the A phase and B phase windings in series. A high voltage capacitor (C) within the surge tester is charged from a high voltage DC supply via the winding inductance (L). Once the capacitor is charged to the desired voltage, a switch is closed. The switch is a thyratron in older surge testers, or an IGBT in modern sets. The energy stored in the capacitor then oscillates back and forth with the winding inductance. The resonant frequency (f) of either the voltage or current waveform is approximately: f = 1/2p(LC) 0.5 If there is no turn fault, there will be a fixed frequency of oscillation. If a turn fault occurs as a result of the short risetime surge imposed on the line end turns, together with weak turn insulation, the inductance of the winding will decrease, and thus the resonant frequency will increase. Thus one looks for the increase in frequency of the voltage surge on an oscilloscope screen as the voltage is gradually increased and as the winding moves from no turn shorts to having a turn short. The increase in frequency is small, typically only a few percent. Such a small increase is difficult to detect. To aid in detecting the frequency shift, modern surge testers digitally capture the resonant waveform at low voltage, where the turn insulation is still intact. The surge voltage is gradually increased by raising the voltage that the capacitor charges to, and triggering the switch after the capacitor has charged up (usually the switch is automatically triggered once per second). If a change in the waveform is noted above a certain voltage, which can be detected by scaling the low voltage stored waveform up to the current applied voltage, then turn insulation puncture has occurred. Older surge test sets were called surge comparison testers. They consisted of two energy storage capacitors, which were connected to two phases. The waveform from each phase is monitored on an analog oscilloscope. The assumption is that the waveform is identical for the two phases. As the voltage is increased, if one of the waveforms changes (increases in frequency) then turn puncture occurred in the phase that changed. This approach has lost favor now since it is possible for two phases to have slightly different inductances due to different circuit ring bus lengths, midwinding equalizer connections or even due to rotor position (since it affects the permeability). It is easy to detect turn insulation failure on individual coils, since the shorting of one turn will have a much larger impact on the total inductance of a coil, thus drastically changing the waveform. Machine manufacturers and rewind companies use individual coil surge testing to check the quality of the turn insulation. Such testing is best done after the coils are wound, wedged and braced since by then they have been exposed to all the mechanical handling and stresses associated with the winding process. B. Test Method IEEE 522 provides the best description of both an acceptance and maintenance surge test for form wound stator windings. The existing standard is being revised to recognize new types of digital surge testers, but the test voltages remain the same. As an acceptance test, the surge is recommended to have a risetime of 100 ns and a maximum magnitude of 3.5 per unit, where 1 per unit is the peak line to ground rated voltage. For a maintenance test performed after the winding has seen service, the surge should have the same risetime, but reach only 2.6 per unit. As for the DC and AC hipot test

10 voltages, these limits were set because they represent the worst surge that is most likely to occur in normal service. Voltages higher than these maximums should not be applied to the stator winding, otherwise there is a significant risk that good turn insulation will fail unnecessarily. As discussed above, the surge voltage is gradually increased to the maximum recommended test voltage. If the waveform changes on the oscilloscope, then the turn insulation has likely punctured. If the winding is form wound, the failed coil will have to be located and isolated. If a turn puncture has occurred, it is not acceptable to ignore it, and return the stator to service. Once the first significant surge occurs in service, the punctured turn insulation will breakdown again, allowing power frequency currents to flow, rapidly leading to groundwall failure. C. Interpretation The surge test is a go-no go test, and the stator either passes or fails. There is no real diagnostic information obtained. If one combines the surge test with a partial discharge test, then it may be possible to detect significant voids between the turns, before actual puncture occurs. This requires a special PD detector, since conventional PD detectors will be damaged by the high voltage surges [12]. VII. IEEE 1434 PARTIAL DISCHARGE TESTING IEEE 1434 is a new diagnostic testing guide that was just issued in 2000 [6]. The guide describes off-line and on-line partial discharge (PD) test methods. A PD test directly measures the pulse currents resulting from PD within a winding. Thus any failure process that creates PD as a symptom can be detected with this method. The test is mainly relevant for form-wound stator windings rated 2300 V and above. As described in IEEE 1434, there are a large number of test methods: 1. Off-line PD test on the entire stator to quantify the PD activity. 2. TVA (corona) probe test to locate the PD. 3. Ultrasonic probe test to locate the PD. 4. Blackout or ultraviolet imaging to locate the PD. 5. On-line PD test to quantify PD during normal service. The first four methods are performed with the motor or generator out of service, and in some state of disassembly. The last test is performed during normal motor or generator operation, however, either an expert must do the test, or advanced electrical interference technology is needed to ensure electrical noise does not cause a false indication. A. Purpose and Theory Many stator winding failure processes exhibit PD as a direct cause or as a symptom of the process. When a partial discharge pulse occurs, there is a very fast flow of electrons from one side of the gas filled void to the other side. Since the electrons are moving close to the speed of light across a small distance, the pulse has a very short duration, typically a few nanoseconds. Since the electrons carry a charge, each individual discharge creates a current pulse. In addition to the electron current flow, there will be a flow of positive ions (created when the electrons are ionized from the gas molecules) in the opposite direction. Each PD pulse current originates in a specific part of a winding. The current will travel along the coil. Since the surge impedance of a coil in a slot is approximately 30 ohms, a voltage pulse will also be created, according to Ohm s law. The current and voltage pulse flows away from the PD site, and some portion of the pulse current and voltage will travel to the stator winding terminals. A Fourier transform of a current pulse generates frequencies up to several hundred megahertz. Any device sensitive to high frequencies can detect the PD pulse currents. In a PD test on complete windings, the most common means of detecting the PD currents is to use a high voltage capacitor connected to the stator terminal. Typical capacitances are 80 pf to 1000 pf. The capacitor is a very high impedance to the high AC voltage (needed to energize the winding sufficiently to create the PD in any voids that may be present), while being a very low impedance to the high frequency PD pulse currents. The output of the high voltage capacitor drives a resistive or inductivecapacitive load. The PD pulse current that passes through the capacitor will create a voltage pulse across the resistor or inductive-capacitive network, which can be displayed on an oscilloscope, frequency spectrum analyzer, or other display device. The bandwidth of the detector is the frequency range of the high voltage detection capacitor in combination with the resistive or inductive-capacitor network load. Early detectors were sensitive to the 10 khz, 100 khz or 1 MHz ranges. Modern detectors can be sensitive up to the several hundred megahertz range [6]. In addition, high frequency current transformers are sometimes installed on surge capacitor grounds to detect the PD [6]. Every PD will create its own pulse. Some PD pulses are larger than others. In general, the magnitude of a particular PD pulse is proportional to the size of the void in which the PD occurred. Consequently the bigger the detected PD pulse, the larger is the defect that originated the discharge. In contrast, smaller defects tend to produce smaller PD pulses. The attraction of the PD test is that one concentrates on the larger pulses, and ignores the smaller pulses. In contrast to the power factor tip-up test, which is a measure of the total PD activity (or the total void content), the PD test enables the