By Howard W Penrose, Ph.D., CMRP
President | MotorDoc LLC
Properly assessing an operating turbine’s electrical and mechanical systems is no easy task. An electrical signature analysis or ESA is independent of load or speed, so the test is often applied to analyze generators and electric equipment. Still, a challenge is to ensure loading stays constant during each test and at the same speed for variable-speed generators.
During an ESA test, rotational speed multiplied by the number of individual components provides useful information. The amplitude of the peaks on FFT spectra (an analysis that transforms a waveform into the components of its frequency spectrum) is also analyzed in relation to the peak voltage and current line frequency. The peak voltage and current changes depending on loading or output and all associated peaks remain relative to the voltage and current peaks.

These graphs represent low-frequency data. The top graph is RMS data, the middle graph is the low frequency FFT, and the bottom graph is the demodulated FFT.
There are few exceptions to this rule and the conditions remain the same even when the amplitude of the peaks change slightly. An ESA also provides information such as power factor, efficiency, phase balance, harmonics, and other important power-quality data that is valuable when performing an analysis. This is significant because a single test or data set alone cannot provide an accurate view of the complete system.
The best approach is a holistic one with an examination of multiple measurements. For instance, information about the machine tested and its controls, such as variable-frequency drives (VFDs) are just as important as voltage and current data.
For a wind turbine, useful pretest information includes a:
- Complete machine nameplate, especially the horsepower or kilowatt, base rpm, voltage, and current
- Number of rotor bars (or slots) and stator slots
- Bearing information
- Type of couplings
- Gearbox data, such as:
– Gear teeth or
– Ratios for each shaft - Bearings and their shafts and locations
- Number of blades for direct-drive pumps and vanes
- Belt-application data, such as:
– Sheave sizes
– Belt information
– Shaft center-to-center distance
– Control information, and - All other components attached electrically or mechanically to the system.
This list of components is important for an accurate analysis. The data collected are then related to the speed of the machine, and multiplied by an appropriate and predetermined number. For example, stator (the stationary part of an electric generator) issues are related to the number of stator slots times the rotational speed in rpm. How this data is finally presented is up to the analyst. Some prefer to use Hertz and others prefer rpm or cpm (cycles per minute). Either method is valid and, ultimately, the more information that’s available, the better.
Basic analysis
Data collection for a generator is either performed using voltage probes and current clamps, or by connecting a preinstalled plug that can jack into a data collector. Because exposed energy is probable and hazardous, it’s imperative to follow safe methods of data collection. Collecting data from current transformers or potential transformers is acceptable but frequency and amplitude are often dampened.
The first step in a generator analysis reviews the applied voltage and current, so ensure both are collected. The RMS (root mean square) and waveform of just the voltage and current will say a lot about how the equipment is operating. It will become apparent if there are problems, such as cyclical loads, VFD issues, or other concerns.

The upper left data is the 0.05 second current, the upper right is the high frequency current FFT, the lower left is the 0.05 second voltage, and the lower right is the high-frequency voltage FFT. Data such as this is often used for tracking harmonic conditions and higher frequency faults.
Higher resolution spectra around the line-frequency voltage and current will relate directly to conditions effecting the rotor and driven equipment. Some bearing, fan, impellor, and gear issues also show up in this range. Normally the data is viewed in decibels (dB) measuring from the peak voltage and current (0 to -100 dB). This provides a relationship in the relative force that’s associated with the evaluated frequency. These peaks mirror each other on either side of the line frequency. Demodulated voltage and current are related to the mirrored peaks.
This data is often presented in voltage and current frequencies without the actual line frequency peak. It can help determine the rotational speed and the peaks to look for in the low-frequency spectra. It is not, however, used for determining alarm conditions. Higher frequency data is used for looking at harmonic conditions and higher frequency faults, such as stator conditions, rotor conditions, and eccentricity. At higher frequencies, points of analysis are formed as twice, line frequency pairs.
Waterfall analysis
Depending on the system, a waterfall analysis – a visual representation of the cumulative effect of data – can provide further insight. This is especially true when frequency peaks have broad bases, which can indicate looseness in a variety of system components, changes in speed, or torsional issues. The waterfall analysis provides a “Z-axis view” such that an analyst can review changes over time.
The front or X-Y axis view provides a layered or average view across the time span when data was collected. When cyclical issues exist, often there’s a rise and fall across the Z-axis. Looseness of powertrain components is indicated with a fluttering across time
Torsional analysis
Torsional vibration is a concern in power transmission systems that use rotating shafts or couplings because it can cause failures when not properly controlled. Torque ripple, an increase or decrease in output torque when the shaft rotates, is measured by calculating the difference in maximum and minimum torque over one complete revolution. This is often difficult to detect with vibration analysis and requires expensive strain gauges mounted on the rotating components.
Fortunately, an ESA lends itself to the direct detection of torque ripple. Such ripple can then detect the associated frequency and the dB down from the peak voltage or current. The frequency, if significant, will also show as peaks in the low-frequency and de-modulated spectra.
Torque ripple will reveal data on the driven equipment, couplings, and even the controls. For instance, a variable frequency drive will have an amplitude of ripple, and with greater severity, can result in coupling faults, shaft failure, or equipment failure.
Evaluating the generator
Just as with induction machines, the transducer for the generator is the air-gap between the rotor and stator. This air gap increases the reluctance between the stator and rotor, which enhances the magnetizing current. Small variations or changes in the magnetic field stem from the rotor vibration and are detectable by ESA. The data rides on the voltage and current waveform, which is an amplitude-modulated signal.
When collecting data on a generator, it’s important to read information from the stator before transformers or other devices are in the circuit because they will block the signature intended for inspection. All existing ESA devices have a 600 Vac limit with the exception of the ALL-TEST Pro OL that has a 1,000 Vac limit. The ALL-TEST Pro OL is a power quality and ESA data collector that has been used on motors, generators, and wind turbines since 2002.
For a system with voltage higher than the limit of the instrument, data is collected using current transformers (CTs) or partial transformers (PTs). However, CTs and PTs will cause some damping and possibly even a small phase shift unless they are specifically selected for ESA testing.
For proper analysis, it’s critical that data for all three phases are recorded in voltage and current. When using CTs or more than one conductor, enter a CT ratio to ensure a proper loading evaluation. Bearing signatures, when present, are a significant find regardless of their amplitude.
It takes a significant amount of damage to the bearing races or balls to cause the rotor to vibrate enough for detection. An ESA provides a secondary method for bearing fault detection. Details of the method for analysis are beyond the scope of this article, but data collection is still worth the time and effort to ensure an optimally performing machine.
Proper analysis requires significant data for an accurate reading. Overall, an ESA can detect conditions through the entire generator power train, from the prime mover to generator load.
For a digital copy of the entire October issue of Windpower Engineering & Development, click here.
Filed Under: Generators