Variations in the load and speed of motors produces corresponding variations in current and voltage, since motors have an innate ability to function as transducers. These electrical changes can be monitored in the time and frequency domains to characterise loads, stresses and wear throughout systems of motors, bearings, transmissions and loads. This can allow a single sensor attached to a cable to provide a wide range of diagnostic information. As with other forms of condition monitoring, early detection of issues can enable preventative maintenance and prevent unplanned downtime or secondary damage.
Because sensors can be located remotely, anywhere on the supply cables, ESA is particularly well-suited to equipment operating in harsh environments. The sensors used to detect electrical perturbances can be safely located in the controller cabinet, removed from the environment where the motor or generator is operating. This is quite different to vibration monitoring where sensors must be in mechanical contact with the body of motors, gear boxes and bearing housings. Applications where this can be a major advantage for ESA include immersion pumps down well shafts or in nuclear power plants.
“With most wind turbines I’ll walk up to the base of the tower, take a one-minute set of data - from the base, not have to climb,” says Howard Penrose, president of MotorDoc.
Signal Condition and Analysis
ESA includes motor current signature analysis (MCSA), voltage signature analysis (VSA), motor circuit analysis (MCA), Extended Park’s Vector Approach (EPVA) and Instantaneous Power Signature Analysis (IPSA). It is based on a range of signal conditioning and signature analysis methods developed by the Oak Ridge National Laboratory. Fast Fourier Transform (FFT) is used to turn time-domain signals into the frequency domain in order to identify frequency signals characteristic of specific faults.
Typically changes in condition or load will be observed as a change in current for a motor and a change in voltage for a generator. MCSA is most sensitive for detecting misalignment or broken bars; EPVA is best for stator electrical imbalance and vibration analysis is best for bearing faults.
Current is typically detected using a transformer attached to a motor’s supply cable, so that a current is induced in the sensor. The current or voltage signal first undergoes some signal conditioning in the time domain, and a spectrum analyser which uses FFT converts this into a frequency domain signal for analysis. Without any other disturbances, the current supplying a motor would be a perfect sinusoidal wave. In reality, electrical and mechanical faults and imperfections result in numerous sideband harmonics. These disturbances combine to produce what appears typically appears to be random noise in a time domain plot. This is the value of a spectral plot, since it allows the frequency and amplitude of each individual disturbance to be viewed on a single chart.
Two types of plots can be used to present the results:
A time domain plot shows the current or voltage on the y-axis against time on the x-axis.A spectral plot shows the signal in the frequency domain, with amplitude on the y-axis against frequency on the x-axis. Spikes on the plot correspond to the frequencies of underlying waveforms. MCSA for Induction Motors
One of the most common implementations of ESA is in monitoring the current supplying induction motors. This enables the detection of electrical faults such as opening or shorting conditions on the stator and mechanical faults such as bearing failures. While it is the current flowing through the stator that is directly monitored, currents induced by the rotor are also indirectly detected. It is typically only necessary to monitor a single phase supplying a motor, a method known as single stator current monitoring.
Motor faults produce disturbances with frequencies typically between 0-5 kHz. The Nyquist theorem shows that accurate signal characterisation requires a sampling rate at least two times the highest frequency being measured. Therefore, the sampling rate should be at least 10 kHz.
Of MCSA, Penrose says: “I generally use it as my system analysis. Because I’m using the magnetic field between the rotor and the stator of a motor or generator of any type, and any size, I can see everything going on downstream in current, all the way to the pump and some of the valves, and whatever else is in that system. I can see the power quality going upstream, sometimes even past the transformer.”
Motors most commonly develop issues due to issues with bearings and stators, and to some extent also with rotors. MCSA is the most effective way to detect stator and rotor faults, and to determine if there is misalignment between the stator and rotor (air-gap eccentricity). Although vibration analysis is sometimes able to detect stator short-circuits in stators and loose or broken rotor bars, MCSA is much more sensitive. Conversely, while it is possible to detect bearing faults with MCSA, vibration analysis is more sensitive to those.
“Because the carrier frequency is the line frequency, you have peak pairs, instead of individual peaks. What’s phenomenal about that is, if I’m a vibration analyst and I go in and have a bearing issue, and I know what the multipliers are from a bearing database” adds the MotorDoc president.
While there have been issues in the past differentiating between power supply voltage waveform distortions and motor faults, newer analytical methods are providing improved diagnostic capability. Examples include model-based voltage and current (MBVI), space vector current modulation (SVCM) and total harmonic distortion (THD).
“It’s best to use vibration analysis for mechanical faults like bearing faults and misalignments, while MCSA has better sensitivity to electrical faults. MCSA and vibration analysis are complementary methods that should be used together when resources permit,” says Dr Dubravko Miljković of the Croatian Electricity Company.
In summary, although ideally condition monitoring will include multiple sensors, there must also be consideration of cost. Vibration sensors are more expensive and need to be located close to the source of vibration. This further increases cost, for example due to the time required for an analyst to climb a wind turbine are install vibration sensors. There may also be issues with operation of vibration sensors in harsh environments. ESA is therefore an ideal form of condition monitoring for many situations. It is ideal for continuous monitoring, harsh conditions or areas with restricted access.