[April 11, 2014] Determining full-load motor temperature from no-load tests [Electrical Apparatus] Tweet (Electrical Apparatus Via Acquire Media NewsEdge) The drawbacks of various testing methods-and some possible solutions EFFICIENCY, TORQUE, power factor-none of these a-c motor performance features means anything if the motor overheats at full load. The ultimate test of any design, then, is the heat run-the proof that rated temperature is not exceeded. The simplest test method is to couple the motor to a dynamometer loading device (Figure 1), apply full rated horsepower, and measure motor winding temperature after it has stabilized. For motors up to 2,500 hp, such testing is possible in many service centers, where it's done to verify that repair methods have not risked overheating (Figure 2). At some motor factories, motors up to 5,000 hp can be "dyne tested" (Figure 3). Still, direct-loading test capability is not available for all ratings (Figure 4, next page). It's the procedure required in the U.S. to verify efficiency of motors subject to Dept, of Energy regulations. However, like any machine, the dynamometer has both speed and torque limits. Power is proportional to the product of the two. Capability of a typical water absorption dynamometer is shown in Figure 5 (next page). What if an adequate loading device isn't available? Several methods are possible to simulate the motor heating that would normally occur at full load. Internationally, some tests were first standardized in 2002 by IEC 61986, titled "Equivalent loading and superposition techniques-indirect testing to determine temperature rise." This standard was superseded in 2008 by IEC 60034-29, applicable to d-c machines as well as both synchronous and induction motors. All these tests attempt to develop heat-producing losses with the motor (usually, but not always) uncoupled from any shaft load that will "approximate" those at rated load. Current, voltage, and magnetic field waveform deviations inherent during the test make an exact match impossible, but the results are commonly considered "close enough" as confirmed by comparative load tests. Although no-load tests avoid the necessity for a mechanically coupled loading device, some methods do require as many as four additional machines of different types and ratings, as well as sophisticated control systems. The dual-frequency test Probably most well-known is the "dual frequency" (or "equivalent load," "synthetic loading," or "artificial loading") test, in which a combination of two power sources at two different frequencies supplies rated full-load current at rated voltage to the motor being tested, with no mechanical load coupled. In the U.S., 60Hz machines are normally tested with 60Hz power from one source and 50 Hz from the othet The two frequencíes are "mixed" either in the winding of the 50 Hz generator (Figure 6), or in a transformer with windings supplied at both frequencies (Figure 7). The second method is more common because a transformer with multiple taps for a range of test voltages is more economical and requires less maintenance than a 50 Hz generator. The 60 Hz "main" source can be the utility line, but a separate generator is the usual choice because that eliminates any feedback of harmonics into the utility system. IEEE Standard 112, Para. 8.2.3, briefly describes the test procedure. The purpose of the two frequencies is to drive the motor rapidly back and forth between motoring and generating modes, with rapidly changing slip, to develop internal losses that closely approximate those at full load. The nonsinusoidal waveform of the applied voltage does cause some deviation in the core and stray losses, but the temperature rise measurement is normally within a few degrees of what would be observed in a dynamometer test. In a 2002 paper, engineers representing one large motor manufacturer contended that direct loading and dual frequency tests on a group of motors from 400 to 4,000 hp yielded "similar" temperature rises; most results were within 2%. Obviously, the test cannot be used to verify efficiency. Vibration will tend to be more severe than during a direct load test, which can be troublesome for vertical motors subject to the "reed frequency" vibration phenomenon. Since the motor on test operates with rated voltage VR and current IR, transformer or generator windings in the 60 Hz circuit must be insulated for that voltage and be capable of carrying that current. The power need only supply the full-load losses in the motor on test. In one example, consider testing of a 1,000 hp 2,300 volt 60 Hz motor with rated full-load current of 225 amperes. Secondary voltage, when 225 amperes is flowing in that winding, would be between 10% and 30% of the 2,300 volt motor rating (figures quoted elsewhere between 1970 and 2009 include 5, 13, 20, 25, and 30 percent). A value of 11%, or 250 volts, was chosen. A typical transformer ratio useful in supplying the 50 Hz frequency would be 460/2,300 volts. Current in its primary would be 2,300/460 times 225 or 1,125 amperes, and the corresponding primary voltage would be (460/2,300)(250) or 50 volts. These current and voltage figures require a 50 Hz source capability of (1.73)(1.125)(50) or only 97 kVA. However, a generator rated to deliver 1,125 amperes would typically be designed with an output voltage rating of at least 460, corresponding to a kVA rating of (1.73)(0.46)(1,125) or 900 kVA. The transformer rating will be at least 1.73)(2.3)(225), or, again, 900 kVA. The 60 Hz source will supply 2,300 volts at 225 amperes-again, 900 kVA. Hence the general rule that the dual frequency test requires both frequency sources and the transformer, if any, to have about the same kVA rating as the motor to be tested. The power required during the test, however, will only need to supply the losses involved in the various components. This is a great advantage for large machines when full-load power would be limited by the local utility-especially since a heat run, unlike many other tests, requires that power usage for several hours. The dual frequency test for replicating full-load temperature rise appears to have first been proposed in a 1921 paper by Arle Ytterberg of ASEA in Sweden. However, another method of using two frequencies had been in use years earlier, as reported by A. M. Dudley in the U.S. Dudley's method involved duplicate machines coupled together, each supplied at a different frequency. The frequencies were adjusted until both motors operated fully loaded, one as a motor, the other as an induction generator. Total power supplied was only that required for machine losses. The three-step test Combining features of other methods described thus far is the "three-step" test, also based originally on IEC standard 61986 (now 60034-29). The three steps are: 1. Apply rated voltage and frequency. Measured temperature rise T1 will be essentially that produced by core loss alone. 2. Simulate rated load by dual frequency operation, but at typically half rated voltage. Temperature rise T2 will be that due to full-load stator and rotor currents, stray load loss, and reduced voltage core loss. 3. Apply about half rated voltage (the same as in Test No. 2) at rated frequency. Temperature rise T3 is that due to reduced stator current and reduced voltage core loss. By superposition, the stator temperature rise at rated operating conditions is taken as Tl + T2 - T3. The superposition principle appears sound. Consider, however, the relationship between stator winding and stator core losses when the machine is operating at normal full load. Temperature of the winding has no influence on the core, but the core temperature does influence the winding. The more the core is heated by its own loss, in close contact with the winding, the more the winding resistance rises, resulting in higher winding temperature than if the core were not present. Thus, adding winding temperature measured with low core loss (Test No. 2) to the core loss temperature rise alone (Test No. 1) doesn't exactly match full-load conditions. Also, as proponents of this method recognize, the assumption is that stray load losses depend only upon stator current. How does the three-step method compare with the "singlestep" dual frequency procedure of IEEE 112? Results of both tests on four large medium-voltage machines showed an average difference of 2.6°C (measured by embedded detector), suggesting that little is gained by adding the extra tests. It's been claimed that "hot vibration ... at the full load condition" can be measured during Test No. 2. But that test is at reduced voltage, meaning the influence on vibration of magnetic pull between stator and rotor will be greatly reduced. Observable vibration will only be that associated with mechanical unbalance and shaft deflection. Is the three-step method "accurate"? Yes or no, depending on the degree of accuracy the circumstances require. Results seem satisfactory for machines in general. The larger the motor, the more likely its design will not be limited by temperature rise. Meeting user needs for low inrush current, severe acceleration, or high efficiency tends to result in cool-running motors. Hence, possible error of 2 or 3 degrees in full-load temperature is almost never significant. Also worth noting from reported three-step tests: wide variations ranging from 4°C to 30°C have been observed between temperature rises measured by resistance and those measured by detector. A 'new' method that's not that new What appears to be a similar method was presented at the 9th MELECON Mediterranean Electrotechnical Conference in Israel in 1998. A 2000 publication offered a "new method of measuring the full load temperature rise of three phase induction motors," in which the motor is run without load at a voltage "slightly higher than the rated voltage." The author claimed that an input voltage "about 120%" of the rating causes "full load losses to occur" with consequent full-load temperature. This method was hardly new. It was briefly mentioned in a 1913 AIEE paper as one of several ways to approximate fullload losses without any direct load. The paper described the procedure as "operating the tested unit as a motor without load but at a voltage higher than normal in the effort to increase the iron and no-load copper losses to a sum approximating their full-load values." What happens is that the machine is driven into magnetic saturation. The normally low magnetizing current becomes much higher, increasing stator I2R loss, while the overvoltage drives up the core loss. Some discrepancies are readily apparent. First, some elements of full-load stray loss are current-dependent, while others are voltage-dependent. If either current or voltage is not the rated value, stray loss will deviate from the full-load value by an uncertain amount. Secondly, regardless of stator current, rotor I2R loss will be low because the slip is low. Some synthetic loading tests briefly described in papers dating back to the I940's refer to superposition of temperature measurements taken at rated frequency but different voltages. In the so-called "copper loss test," temperature was measured "at reduced voltage and full load current at several times normal slip." How this was done was not explained. A second "iron loss test" applied rated voltage to the unloaded motor, and the measured temperature rise from that was then added to the result of the copper loss test. The forward stall test Although the test methods described thus far will yield a close approximation to the normal full-load temperature rise, they do not involve the same magnetic Held conditions as at rated shaft horsepower and voltage. Therefore, normal vibration levels cannot be observed. This is of particular concern for high-speed machines in petrochemical applications. One procedure intended to overcome this limitation has been described as the "forward stall" or "forward short circuit" test, developed in England more than a half century ago. Published results show differences from 1 to 7 degrees compared to full-load testing. Figure 8 illustrates the setup. The Motor M. to be tested, is driven at rated speed by Motor MI, which is commonly a d-c or wound-rotor unit to achieve close control of RPM and is sized at about one-tenth the rating of M. Electrically, M is fed by the generator G, which in turn is driven by Motor M2 at a speed such that the generator output frequency is 80-85 percent of the rated frequency of M. For a 60 Hz machine, then, the generator supplies about 50 Hz. (Note the resemblance to the dual frequency method.) Increasing the generator voltage to about 25% of the rated M voltage causes rated current to flow in M, which operates as an induction generator at high slip. The distribution of individual losses within M will differ from normal full-load operation, but the overall effect is an observed winding temperature rise quite close to the full-load value. Here's what to expect: 1. Stator FR: since stator current is of the rated value, this loss will match full-load. 2. Rotor FR: this will exceed the rated load value because of the reduced frequency and rotor current above the full-load value by a relatively small amount for high-speed machines; a larger amount for low-speed ratings. This is a factor in choosing generator frequency, and its importance must be determined through experience. 3. Core loss: at reduced frequency and voltage, this will be quite low, and have little influence on temperature. From noload heat runs at rated frequency and both rated and reduced voltage, a temperature increment for core loss alone can be obtained and added to the measurement obtained from the forward stall test. 4. Stray load loss: reduced frequency will reduce some of this, but most will see little change. 5. Friction & windage: since M runs at rated speed, these losses will be the same as at rated load. More important: cooling will also be the same. Table I shows comparative values from direct full-load testing and the forward stall method for a 1,500 hp 1,800 RPM motor. Because Motor M is at rated speed, mechanical vibration will be essentially the same as during a full-load test. Harmonics are absent with only a single frequency present. However, as one author has conceded, "the voltage-sensitive portion of the vibration will be lower than normal," because the magnetic flux in the air gap will not match the full-load, full-voltage value. He has suggested checking vibration at different voltage levels to determine if the "electrical content" is "significant"-although giving no guidelines for making the judgment. Here again, despite the extra machines and interconnections required (both electrical and mechanical), the power to the machine on test need only supply motor losses rather than the full power rating. The synthetic loading method A still more complex procedure was the "new synthetic loading method" diagrammed in Figure 9, proposed in 2002 by Canadian engineers who claimed the dual frequency test was subject to "major inaccuracies." In a typical arrangement, motors Ml, M2, and M5 are squirrel-cage motors; Gl and G2 synchronous generators; the two generators are mechanically coupled. Circuitry for the generators controls power swings between system units to require from the utility source only enough power for machine losses while generating essentially full-load heating within the test motor. Details are too complex to include here, and any commercial use of this system for large motor testing has not been publicized. During the 1990's, researchers at Australia's University of New South Wales developed "a novel method for rapid efficiency measurement of three phase induction motors." It was originally aimed at evaluating motor full-load temperature rise without the high power supply required in a dynamometer test. Two possible approaches were reported. The first involved the use of microprocessor-controlled electronics to create two power sources at different frequencies. One shortcoming was recognized: . . the above analysis . . . does not include the stray losses . . . due to space and time harmonics which are inherently present when induction motors are supplied from an inverter...." In a related "sweep frequency" test, an electronic power supply rapidly modulates a single supply frequency between two limits. In a typical test of a 60 Hz motor, the "sweep frequency" might be varied from 50 to 70 Hz at a rate of five times per second. An alternative procedure, called the "sweep frequency" method, used microprocessor control to rapidly modulate a single supply frequency. Earlier reports in the U.S. had pointed out that the dual frequency test "can" produce dynamic (oscillating) torques reaching nearly 1-1/2 times the full-load torque rating of the motor on test. The sweep frequency test was found to create even higher levels of torque oscillation. This gave rise to high torsional vibration, which is one of the drawbacks of any mixed-frequency testing of vertical motors that are normally attached to a firm base at only one end. Either of these Digital Signal Processing (DSP) methods will cause the motor on test to alternate rapidly between motor and generator operation. The losses can be adjusted to their full-load values, just as described earlier here for the conventional dualfrequency test. Refinements of the Australian testing were said to allow efficiency measurement as well as temperature. Some limitations remain, however. One is that the non-sinusoidal voltage waveform and associated magnetic flux distortion result in deviation in core loss and stray load loss. Testing of a 10 hp motor produced these results: Leaving aside the question of whether results can claim to be accurate within 1 watt in nearly 10,000, agreement among the efficiency values is described as "close." It is, however, not close enough to serve as verification of U.S. Dept, of Energymandated efficiency levels, and DSP methods are not recognized for that purpose. They do remain useful for temperature measurement. More important: Applying DSP tests to motors ranging up into the thousands of horsepower, at medium voltage, requires extremely large and costly electronic gear, with the ready capability of supplying several different voltages up to 13 kV to suit different motors. There is no indication that DSP testing has been widely adopted. In 1972, Chinese researchers proposed a "phantom loading" method in which a single a-c voltage and frequency is applied to the paralleled windings of two motors. The other ends of those paralleled windings is connected to a "variable d-c supply" (see Figure 10). This test has not become widely used for a number of reasons. First, for wye-connected motors, both ends of each phase must be brought out, which is not common for large machines. Second, the test cannot be performed on a single motor. Third, the d-c supply rating must be determined, for which the proposal offered no guidelines. Limits of estimating full-load temperature rise Several other ways of estimating full-load temperature rise can be useful when a direct-connected load is available but at less than the full-load motor rating. Data is extrapolated from measurements at partial load. Two assumptions may be useful under certain conditions, but not always. The first is that winding temperature rise always varies directly as the square of the load current. Another, related to it, is that temperature rise necessarily varies directly with horsepower. Figure 11 shows the typical relationship between load and temperature rise. For a motor operating at or near full load, an increase in horsepower output P will typically result in an increase in temperature rise equal to P16. That's because the rise varies as the sum of the heat-producing losses in the machine. Each loss, however, does not equally affect temperature. Copper loss in the stator has an influence typically measurable by 1.15 times its value, core loss by 0.75 times its value, and the rotor and stray losses at 1.0 times. At low loads, the effects are different. Core loss predominates, tending to reduce the influence of the other losses. In ratioing up from about one-half load to full load, then, temperature rise will more nearly vary directly as the change in load. More accurate calculations would of course require knowing what all the losses are at a given load. This is rarely true outside the factory. A possible solution One possible way to deal with this shortcoming is with a series of tests based on the following: Motor internal losses at rated voltage fall into two categories. The fixed losses are essentially independent of power output. They include core loss (CL) and friction & windage (FW). The variable losses, considered directly proportional to line current squared, include stator I2R, rotor I2R, and stray. Because each of these two groups will contribute to winding heating, their effect may be added to arrive at total temperature. Since all three variable losses are related to current in the same way, only the variation in stator I2R need be considered, so that this equation can be written as: Degrees C rise = (fixed loss in kW)(A) + (variable loss in kW)(B) in which A and B = degrees C per kW loss, for each category The motor to be tested is first run uncoupled. Winding temperature is measured. Input power is also measured, at several different voltages, so that core loss can be determined. Next, the maximum available load is applied for two testsone at rated voltage, the other at a reduced voltage. Again, winding temperatures are recorded. From the equations just given, the coefficients A and B can then be determined. If we know what all the motor losses should be at full load, the full load temperature rise can be accurately calculated. As we'll see, however, for a typically large difference between the maximum load applied and the full-load rating, such precision isn't necessary. Here's an example taken from one manufacturer's practice: Motor rating 3,000 hp 4 pole, 4,160 volts. Maximum available dynamometer rating is 1,500 hp. Five tests are involved. The first is no-load saturation, in which input power to the unloaded motor is plotted vs. the square of the applied voltage. Extrapolating this plot to zero volts yields the value of FW ( 13.7 kW), and subtracting that from total power input gives the core loss (very nearly) of 5.7 kW at 2300 volts and 22 kW at 4160. Next are successive no-load heat runs at 2,300 and 4,160 volts. The results: 6.6°C rise at 2,300 volts = 5.7A + 0.125B + 13.7C, in which the thermal coefficients are A for the core loss, B for stator copper loss, and C for FW (all in units of degrees C per kilowatt). At 4,160 volts, 16.9°C rise = 22A + 0.54B + 13.7C. If the small B terms are neglected, the two equations can be solved for the values of A and C, from which A = 0.63 degrees per kW and C = 0.22 degrees per kW. (As we'll see, the ultimate result will confirm that FW is not a significant contributor to motor temperature.) The final two tests are heat runs with maximum available dyne load of 1,435 hp. The results: At 2,300 volts: 25.1°C = 5.7A + 7.5B + 13.7C At 4,160 volts: 20.9°C = 22A + 2.6B + 13.7C Comparing the equations for the two 4,160 volt tests, we can solve for B, which = 2.5°C per kilowatt of stator loss. Without going through the rest of the mathematical detail, the final values of A, B, and C can be applied to the tested CL and FW amounts to estimate the temperature rise at the calculated full-load stator I2R of 11 kW. It amounts to 12.1 + 27.3 + 4.11 or 43.5°C. The smallest component, 4.11, is contributed by the FW loss. (Most motor designers would consider even that much unlikely.) The more probable result of 40°C is obtained by plotting the temperature test points as shown in Figure 12. (Again, the assumption is that temperature rise varies directly with stator I2R.) If we did know all the motor losses (based on more detailed testing) at 1,435 hp and 4,160 volts, and weighted them as described earlier here, we would get this breakdown of loss kilowatts as they affect temperature rise: Stator copper (2.57)( 1.15) = 2.95 Rotor copper loss = 1.9 Core loss = (22)(.75) = 16.5 Stray loss = 5 Ignoring FW, this is a total of 26.4 kilowatts, with a measured temperature rise of 20.9° from the test at reduced load and rated voltage. Projecting this to 3,000 hp in the ratio of (rated full-load current/test current) squared, we get this: Stator copper loss = (362/182)2)(1.15)(2.95) = 12.9 kilowatts Rotor copper loss = (362/182)2)( 1.9) = 7.2 Core loss = 16.5 Stray loss = (362/186)2 = 19 for a total weighted loss of 55.6 kW. Expected temperature at full load is then (55.6)/26.4)(20.9) = 43°C. If we don't know individual losses at all, simply ratioing temperature rise in proportion to load, based on Figure 11, yields (3,000/1,435)(20.9) or 43.6°C. All these answers are within four degrees of one another. Choosing among possible "synthetic loading" tests depends upon the auxiliary machines, controls, and power sources available. It also depends upon the importance of an "accurate" answer. Finally, keep in mind that several degrees of temperature variation can be expected from the same test-no matter how accurate-on successive motors of "duplicate" design. * By Richard L. Nailen, P.E., EA Engineering Editor (c) 2014 Barks Publications [ Back To TMCnet.com's Homepage ]