|Online condition monitoring and diagnosis can lead to extended operating times
Early detection with online condition monitoring can help mitigate the effects of increased thermomechanical stresses on power plant core components due to medium and peak load operation, writes Frank Ewert
The influence of renewable energy – especially wind energy – is increasing. Consequently, medium and peak load operation of fossil-fueled power for the stabilization of the electrical grid is increasingly required. Especially for large fossil fuel plants this medium and peak load operation results in increased thermomechanical stress on core components like the generator due to the increasing number of load cycles.
This typically results in accelerated aging of the machines, and risk of damage and unexpected failures. As a preventive measure, inspections at more frequent intervals offer insight into the condition of the generator, but cost the power plant expensive outage time.
To minimize loss of revenue, it is helpful to use online condition monitoring and diagnosis systems to receive information about the state of the machine during operation. This will lead to condition-based maintenance with extended operating periods between standstills.
Generator monitoring overview
Generators are in service in power plants for many years. Operation and aging can gradually cause damage to their high-voltage insulation. If early detection of changes in the components is possible through in-process, long-term diagnosis, unscheduled and expensive outages can be prevented and measures can be scheduled and taken to extend the service life of generators.
Monitoring systems for different parameters are available, but at Siemens the focus is on the monitoring of partial discharges (PD), end winding vibrations (EWV) and interturn short circuit (ISC). Other systems for shaft voltage/current and fibre-optic stator and rotor temperature monitoring are available or in development. The typical Siemens generator monitoring concept of the main monitoring systems mentioned above is illustrated in Figure 1 (page 52).
|Figure 1. Overview of SIEMONplus monitoring system|
Monitoring usually starts with the sensors installed inside the generator (vibration and flux) or in the direct vicinity of the generator – for example, inside the iso-phase bus (partial discharge). The sensor signals are routed to a data acquisition unit (AU), located in the turbine hall. All relevant data is stored on the AU and can be downloaded for diagnosis. By extending the system with a central server, up to 20 generators can be monitored and the data can be stored, visualized and compared. The server can be easily connected to any I&C systems, such as a plant information (PI) system or a T3000 system, or to a superordinated Siemens system like the WIN_TS. Once connected to one of these systems, the data can be automatically transferred to one of the Siemens Power Diagnostic Centres, where the data is observed permanently by sophisticated rule-based analyses which include automatic comparison with all relevant and existing generator parameters. In case of any deviations from normal behaviour, Siemens experts are mobilized to perform a detailed evaluation of the data and to provide recommendations for the customer.
Partial discharge (PD) monitoring
Electrical faults in high-voltage components like a turbine generator do not occur suddenly. In nearly all cases, however, defects are announced by PDs bridging part of the high-voltage insulation, and are detectable with RF-measurement methods.
The distribution of the discharges with respect to the phase angle of the generator voltage is characteristic for the cause of discharges, and allows for conclusions about the grade of risk for further operation of the high-voltage equipment.
The PD measuring system can be normalized according to IEC/TS 60034-27-2. By transformation of the measured voltage signals to charge units, being a measure for the transferred electrical energy, an estimation of the risk is possible. Although the generator, as the most expensive electrical component, is generally the main target of supervision, the complete high-voltage area, including the main and auxiliary transformers, is monitored through registration of PD signals.
For the detection of PDs, special coupling capacitors need to be installed in existing voltage transformer cabinets, generator terminal boxes, inside the isolated phase bus (IPB), inside generator neutral cabinets or at other suitable places in the high-voltage area of the power plant.
The coupling capacitors are used to pick up the PD signals from the high-voltage line. The capacitors usually have a capacity in the range of 1oC-9oC. Coupling capacitors with lower capacities are not used due to a decreased sensitivity which does not fulfil Siemens’ requirements. The capacitors are usually equipped with integrated over voltage protection, and have an insulated signal output to avoid any eddy currents along the cable shields.
PD analysis tools
The most well-known and powerful tool is analysis of the Phase Resolved Partial Discharge (PRPD) patterns. These typically show the PD distribution map of PD magnitude vs AC cycle phase position, for visualization of the PD behaviour during a predefined measuring time. Classification and interpretation of these patterns can be accomplished using the international pattern catalogue (IEC/TS 60034-27-2).
Monitoring of a generator usually starts with a fingerprint measurement, which is the first measurement after the monitoring system is commissioned and the generator is at base load. The fingerprint measurement is used as a reference for any future analyses. It further serves to identify narrow-band interference, and to eliminate it with digital interference suppression in order to improve the measurement sensitivity.
Once the fingerprint measurement is done, the trending of the following common PD parameters starts:
- QIEC [oC] – Apparent charge according to IEC60270;
- Qmax [oC] – Maximum charge;
- N [1/s] PD rate;
- QR [oCà‚²/s] – Quadratic charge rate.
According to IEC 60027-2 , the abovementioned common PD parameters should be correlated with the following generator parameters:
- Slot tempratures;
- Cold and warm gas temperatures;
- Stator currents of all three phases;
- Stator voltages of all three phases;
- Active and reactive power;
- Exciter current.
As a basis for Time of Flight (ToF) measurement for localization, it becomes important to evaluate high-resolution oscillograms taken with a sampling frequency of up 125MS/s of four channels simultaneously over one AC cycle of the high voltage (50 Hz 20 ms).
Furthermore, the oscillograms can be used to determine the slew rates of the PD impulses in order to gain information about the source or origin of any PD impulses. For example, the slew rate of PD impulses coming from inside the insulation system, which travel through the generator winding, are damped much more than impulses coming from outside the generator, e.g., the impulses from an IPB supporter with a poor contact to the lead.
Most of the PD monitoring systems on the market provide the possibility of performing ToF measurements for fault localization.
Based on a high bandwidth and high sample rates, the signal behaviour of PD pulses can be analyzed as function of place. The signal shape of the PD pulse and comparison of both test points give some knowledge about the discharge source.
The difference between the propagation times of the measured pulses can be determined, and the possible region of the corresponding PD activity can be restricted. In this way the pulses’ travel direction can be directly calculated.
If there are well-defined signal propagation paths between two test points (e.g., along the IPB), analysis of the travel time allows calculation of the distance of the PC source from both test points. For time differences below the calibrated travel time between test points, the location of the PD source is between the test points and can be determined with an accuracy that depends on the steepness of the pulses and the geometry of the propagation paths.
In general, PD measured in power plants is superimposed with a lot of noise signals, which must be eliminated prior to any further processing. Those signals can be principally divided into sinusoidal and pulse shaped signals.
Sinusoidal noise, e.g., transmission signals on the power lines, often fully masks any PD pulses contained in the measured signal. Because each data set is highly resolved in the time domain, it can be filtered digitally by transforming it to the frequency domain, filtering the resulting signal from the dominating resonance frequencies using digital notch filters and transforming it back to the time domain.
Typical pulse-shaped signals are, for example, the six equidistant commutation impulses of the static excitation system (six-pulses bridge). These impulses are caused by semiconductor switching in the static exciter power converter and are typical in the operation of generators with static excitation systems with slip rings.
These high-frequency pulses couple from the generator rotor winding via air gap into the stator winding, and represent a normal phenomenon in all generators with static excitation. They can be used as a sensitivity check for PD online monitoring systems, which operate in wide-band mode at low frequency range acc. IEC 60034-27-2.
Slot discharges typically appear between the outer corona protection (OCP) and the stator core. The typical appearance of slot discharges is an asymmetrical distribution of the discharges covering the amplitude and number within the two half-waves.
In the negative half-wave, higher PDs appear. In both half-waves the PDs appear between the zero crossing and the maximum/minimum of the high-voltage cycle. The PRPD patterns of slot discharges typically show a triangular shape.
|Screenshot of SIEMONplus EWV monitoring software|
All forces that act periodically cause elastic structures to vibrate. Such forces occur in every generator. Vibration represents a cyclic load for the affected components, and increased vibration levels mean an increased load (the load being proportional to the vibration amplitude). High levels of vibration cause the end-winding assembly to loosen, and can lead to rubbing and fracture of the affected components.
The end-winding sections of generator are particularly excited by the core which vibrates at twice the line frequency; current forces; and bearing and shaft vibration.
The unusually high currents and torques associated with line short-circuits, lightning strikes and out-of-phase synchronization have a particularly pronounced effect on generator end windings. Although generators may not be damaged by such events, it cannot be completely ruled out that such events can cause a certain amount of damage to the stator winding. To prevent any secondary damage, the affected generator must be inspected to identify and repair potentially loose braces or ties, particularly in the end-winding section.
The condition of generator end windings is typically examined visually. Loosening of end-winding assemblies causes secondary damage and increases the cost and effort of any repair measures. Omitted or late repair can mean that the generator has to be completely re-wound. End-winding vibration monitoring thus is a useful tool for evaluating the condition of generator end windings, and also minimizes the risk of a re-wind.
Online monitoring of end winding vibrations is performed by recording local accelerations during regular operation. Special fibre-optic accelerometers are placed on the bar end connections of the stator end windings (see Figure 2, page 52) considering the results of an offline modal analysis (bump test). These sensors are free of any metal parts and do not interfere with the electromagnetic field in the end winding area. A minimum of six, but better even eight, sensors per end winding should be used to allow a reliable data analysis.
|Figure 2. Application of fibre-optic accelerometers inside a stator end winding basket. The fibre-optic accelerometers are non-metallic and do not interfere with the electromagnetic field in the end winding area|
Real-time signals are usually used to check the general functionality of the sensors. The signals of up to 16 fibre-optic sensors are acquired with a sampling frequency of 9 kS/s simultaneously to enable the possibility of performing online modal analyses.
By using a fast Fourier transformation (FFT), the time-domain signals are transformed into the frequency domain. Here the first and second harmonics of the line voltage are usually observed, and the respective amplitudes, which usually represent the vibration level in [à‚µm], are trended separately for every sensor signal. The trend of the vibration values can be displayed and correlated to operational parameters: for example, active power, reactive power and exciter current, which may have a direct influence on the end winding vibration behaviour.
Modern end winding vibration monitoring systems provide an online modal analysis of the end winding vibration signals. The Siemens system provides online evaluation and visualization of the standstill and rotating vibration modes separately. The respective changes in these vibration modes during operation are trended, in order to detect possible changes in the mechanical behaviour of the generator end winding. For such diagnostics a detailed knowledge of the generator design is necessary in order to define potential measures.
In most cases end winding vibrations are indicated by friction dust caused by relative movements of different end winding components (bandings, blocking elements, etc). Permanent mechanical stresses on single bars can lead to fatigue cracks in single strands. In the case of cracked single bars, the heat produced from the cyclic interrupted current leads to perforation and discolouration of the bar insulation.
A major portion of rotor faults are related to some kind of winding fault, such as earth faults, shorted windings, or cracked conductors. Some winding faults do not directly lead to a trip of the generator, but may lead to increased shaft vibrations, excessive wear or simply reduced efficiency. Therefore a rotor winding monitor is a useful tool for fault analysis.
Different mechanisms can be responsible for shorted rotor turns. Most shorted turns are caused by some kind of relative movement in the rotor turns. This relative movement may lead to misalignment of the insulation layer between individual turns, failure of turn-to-turn insulation, or loosening of blocking elements in the end winding region. Relative movement of rotor winding parts is unavoidable due to the fact that the copper of the windings has a thermal expansion coefficient different from that of the insulation materials and the rotor body. During each startup and shutdown the copper windings move relative to the rotor slots and the insulating parts. An experienced and reliable rotor winding design takes care of this thermomechanical mechanism – otherwise relative movement leads to increased wear on the insulating materials, and ultimately to interturn short circuits. Sometimes even plastic deformation can occur, leading to end turn elongation, which in turn may cause the loosening of blocking materials in the end winding region and, ultimately, shorted turns.
The effects of shorted turns depend on their number as well as their location. In the first place, shorted turns require a higher field current to run at a specific load, thereby decreasing their efficiency. Higher field currents result in higher overall field operating temperatures. In the case of a two-pole rotor, shorted turns in one pole lead to increased temperatures on that side of the rotor only. This leads to thermal unbalance in the rotor and increased vibrations due to thermal bowing. Four-pole rotors may suffer from magnetic unbalance in the case of shorted turns in one pole. Interturn short circuits or earth faults in the end winding region may produce arcing underneath the retaining ring, leading to retaining ring damage.
For the detection of shorted turns, a sensor which measures the magnetic flux must be installed in the air gap between the rotor and the stator. The sensor is typically installed in the 3 o’clock or 9 o’clock position, and is fastened on top of a stator slot wedge near the turbine end of the generator. During the unit’s operation the monitoring system analyzes the magnetic flux continuously by detecting changes in the rotor slot leakage flux signal provided by the sensor. It will detect most of the shorted turns which can occur at a generator rotor, and will issue an alarm if a shorted turn is detected. Identification of the affected slot of the rotor is also possible.
The measured magnetic flux does not cross the air gap to reach the stator windings. Its magnitude is proportional to the current flowing through the active turns in each slot. This fact is used to diagnose shorted turns.
Sensitivity in detecting shorted turns in the coils depends on the load point. Therefore it is necessary to use a very sensitive data acquisition system. To detect even the smallest changes in magnitude, the Siemens system uses a sampling frequency of 100 kHz and a resolution of 24 bit. When using such a high resolution, it is no longer necessary to approach different load points between no-load and full-load (flux density zero crossing). Through a pole-to-pole comparison of the heights of the flux probe waveform peaks, the decrease in the number of active turns in the affected pole can be detected.
By trending the data in the case of an existing short, rotor winding defects can be detected in an early stage, and planning for the next maintenance outage can be supported to minimize loss of revenue.
Typical reasons for rotor interturn short circuits include, for example, end-turn distortions or copper chloride contamination. End-turn distortions usually appear due to aging caused by relative movements in the rotor turns. The resulting misalignment of the insulation layers causes failures of the turn-to-turn insulation. Copper chloride contamination can lead to conductive bridges between the turns (see Figure 3, page 54).
|Figure 3. Typical ISC examples|
The more monitoring modules and operational parameters are available, the more successful the diagnostic. Monitoring can help to optimize outages with early recommendations based on the condition of the generator. This leads to minimized outage times, as all necessary repair measures are planned in advance and all spare parts are available in time.
Another goal of generator monitoring is to extend periods between maintenance outages based on experienced analysis of operational conditions, machine characteristics, maintenance history and online diagnostics. And a lifetime extension of generator components such as stator or rotor windings, bushings or circuit rings can be achieved through assisting the maintenance and overhaul of these components.
Frank Ewert is an advisory expert at Siemens Energy. He is responsible for online generator diagnostics as well as generator diagnostic-related research and development projects.
This article is based on a Best Paper Awards winner at POWER-GEN Europe 2014.
Power Engineering International Archives
View Power Generation Articles on PennEnergy.com