EPACT and Motor Testing
Understanding IEEE 112 Method B (or CSA C390)

The goal of all of EPACT (Energy Policy Act) legislation is, of course, energy savings. This is to be accomplished by improving energy efficiency or using less energy to perform the same task. To demonstrate or prove a product's energy efficiency, in this case an electric motor, obviously requires a method of testing. And to compare one product to another requires that this method of testing be standardized, preferably to some recognized national or international standard. Discussions about how EPACT applies to electric motors, presently three phase industrial motors, therefore have always included mention of test standard IEEE 112B, or the Canadian Standards Association's C390, which can be considered equivalent. But what are these standards and what do they mean to the average user?

To understand this we must first realize that there are a number of ways to determine an electric motor's efficiency. The most obvious way, or so it would seem, is to connect the motor to a known load and measure the electrical power into the motor. Assuming we now know both the power (or work) output and power going in, the ratio of these is efficiency. A mathematical way of stating this is shown below:

efficiency	=	output input	=	output output + losses
		or	=	input - losses output + losses	=	fraction of the total input power that produces work or output
(The value of efficiency is then normally converted from a decimal fraction to a % for convenience.)

Most of these losses produce the heat given off by the motor during operation. This energy is therefore not available to perform work.

Motor Losses

Induction motor losses are normally broken into these categories :

• Iron or steel losses: energy lost in the magnetizing of the steel laminations, and to keep them magnetized.
• Stator copper losses: heat generated due to the resistance of the wire as the current flows through it.
• Rotor copper losses: heat generated due to the resistance in the rotor conductors or bars and end rings as current also flows through them.
• Friction and windage losses: energy lost in bearing friction, energy needed to turn the cooling fans and windage of other rotating parts
• Stray losses: other energy lost that does not directly fall into one of the above categories. These are related to the construction of the motor, parts that don't produce output power in or near the magnetic fields in the motor, and interaction of magnetic fields in the motor.

Motor Testing

So why not simply test the motor, as mentioned, to determine the efficiency and total losses? First consider that the efficiency of an electric motor changes as the grease "breaks in" (warms and flows), as the motor materials heat up, and so on. Therefore a procedure must be established to define which efficiency measurement will be considered the "real" or "steady state" efficiency. Secondly, there is the accuracy of the measurements to consider. The output and input power are relatively large numbers that differ by as much as 15% but as little as 4 to 5%. For lower efficiency motors, a slight error in measurement would have a relatively smaller effect on the efficiency. However, with higher efficiency motors (those for EPACT, higher horsepower motors, etc.) where cost decisions are based on only a few tenths of a percent difference in efficiency, great accuracy is critical. Accuracy is another issue that must be addressed in a standard way to get consistent results. Third, a rotating motor and load constitute a "dynamic" system.

Readings of speed, torque, volts, amperes, watts, and temperature are not steady or constant values. There are fluctuations, though small, that must be dealt with during testing in order to be consistent. Basing motor efficiency on essentially one reading could be misleading.

To address all these concerns and more, IEEE (the Institute of Electrical and Electronics Engineers) set out to write standards to define how best to test electric motors. In IEEE standard 112, several methods are described. Efficiency determination is only part of this standard, although it is an important one. Some of the key (there are a total of 10) test methods for efficiency are:

• Method A: simple input-output
• Method B: input-output with loss segregation (or separation)
• Method F: equivalent circuit (model) calculation

The other methods, C, E, E1, F1, C/F, E/F and E1/F1 are variations of these. Very early discussions about efficiency testing concluded that simple input-output was not accurate enough. Method F, using an equivalent circuit approach was considered an indirect determination rather than a direct measurement. There was also a need, by motor design engineers, to know how the total losses were distributed among the various types or categories. This information would allow them to determine how best to improve efficiency and where to focus their efforts.

Of all the methods outlined, only one, Method B, measures input and output and attempts to determine and separate each type of loss. Because of this IEEE 112 Method B became a popular tool and is commonly used in the U.S. motor industry (and in Canada by way of standard C390). Further work has improved the accuracy and repeatability of this method over the years. Experience with the method and this additional work have led to its adoption as the standard and most accurate method for determining motor efficiency.

How Does Method B Relate to the "Real World"?

Testing for efficiency to IEEE 112 Method B is broken down into sections or types of testing. These are no load testing, temperature testing, and testing under load.

During no load testing the motor is connected to rated voltage and frequency and allowed to run without a connected load until the input watt readings stabilize. This can take from over one to perhaps four hours to occur. Once the motor has "loosened up," or stabilized, testing proceeds by adjusting the motor voltage and taking a series of readings from approximately 125% of rated voltage down to a minimum voltage where motor current no longer continues to drop with voltage. Using these readings -- combined with winding and ambient temperatures and winding resistance -- two of the motor losses, iron loss and friction loss, can be determined. The iron loss will vary with voltage while the friction loss will stay relatively constant because motor speed is constant (within a few RPM). The process involves first subtracting out the stator copper loss that can be calculated from the current and resistance (we are not interested in these at this time) and plotting (or mathematically curve-fitting) the remainder. By extrapolating the low voltage data to zero, the constant loss (friction) can be determined. Subtracting this out of the higher voltage readings will allow the iron loss to be calculated for exactly rated voltage.

The temperature testing (or full load heat run) simply involves testing the motor under rated load conditions at rated voltage and frequency and monitoring temperature. The specifications define how precise the load and input voltage must be held. This test must continue until the motor temperature is stable (thermal equilibrium). This is defined as when the motor temperature rise does not change more than 1°C over a period of 30 minutes

Testing under load is done by maintaining rated voltage and frequency while applying six different loads to the motor in one-fourth rated load increments from approximately a quarter load to 1.5 times rated load. Readings such as current, torque, RPM, and temperature will provide information about how the motor performs under load. At this point motor testing is complete and analysis begins.

Known or directly calculated losses are determined by tabulating the readings, making any adjustments to these readings so they are at the correct operating temperature (from the temperature test), and determining each individual loss for each of the six load readings mentioned previously. The iron loss and friction losses were determined in the no-load test. These are held constant for all load points. The winding resistance, corrected for temperature, and measured current will yield the stator copper loss.

Knowing the motor speed (and therefore slip), power input, and other losses at each load will allow the rotor copper loss to be calculated. All of these losses are therefore directly determined or calculated from measured values. However, there is one remaining loss category that has not been addressed, that being stray loss.

Stray loss cannot be calculated directly from an equation based on input or output readings. It is actually the sum of several smaller losses that are dependent primarily on motor (or part) geometry and variation. For the purposes of this test method, stray loss is determined indirectly by subtracting the directly calculated losses and output power from the input power. Because all power (or energy) must be accounted for, this difference is considered to be the stray loss. (There are special tests designed to "measure" or isolate stray loss, but the added time and complexity of testing would not improve the accuracy of this method.) The value of stray loss is typically in the 1% range of motor output power. Even with very accurate measurements, the calculated value of stray loss can vary significantly from reading to reading. This is compounded by the fact that this loss being "what's left" will include small errors or inaccuracies in the measurements or calculation of other losses. This presents a problem.

This entire test method is based on the premise that all losses follow some smooth function of motor load, with other parameters held constant. This potential variation or error often will not allow the stray values to fall on a smooth curve. To address this, the standard outlines a "smoothing" calculation where these remainder values (stray losses) are fit to an agreed upon equation form based on theory. The equation must go through zero, have a positive slope, and the value of stray loss is to vary by the value of torque squared. This smoothing is also intended to "improve" accuracy under the assumption that the true value of stray loss should be closer to values calculated from this smooth equation than the actual test difference values. Therefore, the values calculated must "closely" fit this curve. Specifically, to determine if the readings are correct and/or the test is valid, the values of stray loss must fit this type of curve with a correlation coefficient of 0.9 (90%) or better. If not, there is an allowance provided that a maximum of one point can be ignored if necessary to bring the correlation coefficient up to this level (thus allowing for the possibility of one bad test point). But if this still does not improve the correlation coefficient enough, the test must be taken over.

The calculation of efficiency is done by applying the equation we started with. The efficiency value is determined by first subtracting the losses, now corrected to the operating temperature and smoothed in the case of stray loss, from the known input power to get the corrected output power. Then this corrected output power divided by input power is efficiency as defined by IEEE 112B.

Testing Time, Accuracy, Comparison with Other Methods

The first thing one may notice from this discussion is that these methods require much time, very accurate measurements, as well as patience and experience. Including the preparation of the motor (adding thermocouples to measure temperature, mounting it to a dynamometer of some sort for loading,and hooking up leads and meters), taking readings, waiting for temperatures to stabilize, and performing calculations these tests easily take eight hours or more. The standard defines the accuracy of all measurement equipment to the point of also including a dynamometer correction calculation (not discussed here) to check and account for small errors in output measurement resulting from friction in the dynamometer or other factors. Even so, one can see that there are a relatively large number of measurement calculations which do result in some "calculation error." This includes round off, slight errors in temperature corrections, stray smoothing, and simply error compounding as measurement values are multiplied, squared, and so on. And, one cannot ignore the "human" or procedural errors and limitations in measuring a dynamic system.

NEMA, IEEE, and CSA have worked to address these issues over time with updates to the standard. The latest improvements dealt with tightening up the equipment accuracy and more clearly defining the procedure of testing to minimize "human error." But even with these improvements, variation does exist. A report, published by NEMA, on a "round robin" test program involving several different motors, shows that the variation in total losses on the same motor tested at different facilities, all with proper equipment and using the best practices is still about 10%. Even without a study you can show that even with ± 0.2% instruments and ± 1 RPM speed measurements (both as required by the latest standard) the resulting change in losses can be 4-25% and the range in efficiency calculated for any single load point is almost 1% (0.8-0.9%). See chart above for an example.

It should be noted that one of the original considerations in the development of IEEE 112 Method B was to reduce errors due to measurement limitations that existed in older test equipment. Today, with more accurate equipment available, some question if this method is still the best for efficiency determination. But, in spite of the apparent shortcomings, most feel that it is still the best method of testing and determining efficiency for 3 phase AC motors. It is certainly standardized, widely used in the North America for motor engineering purposes, and perhaps studied the most. It also segregates individual losses. And, if followed properly, the method will be a fair comparison of one motor to another. So it is not likely we will see a change any time soon.

It does NOT however allow for easy verification by the average user. And, one must understand that the tested efficiency is NOT an exact single static number. Although the efficiency calculated from a single rated load input-output reading should come close to the IEEE 112B value, it will seldom be the same. True verification will require a full test. And as we just found out, that too has some variability.

Date 6/19/97 - Technical Note 97-1