Predictive maintenance of high-voltage generators

 

The purpose of this site is to discuss aspects of the predictive maintenance of high-voltage generators. It is written by someone trying to grasp issues and identify the pros and cons of different techniques in terms of return on investment and technical merits of the different classical and novel methods that are presently implemented and marketed to prevent forced outages of generators with medium-to-large ratings.
It refers to different documents collected from different sources either traditional ones or from the Internet.
 I hope the contents of this page will get progressively smarter as I get more acquainted with this field. You are welcome to contribute to this endeavor. If you don't agree with some points, why not mail me.

Some docs and pdfs are pass-word protected.
To have access to them mail me to get the list of passwords in
File with keys
Passwords may differ according to the accessed doc or pdf.
I tried each time to pay a tribute to the companies or people whose contributions appear on this page. Pleas let me know if this is not the case. I'll be happy to take corrective actions.

CIGRE documents

Plant operator documents


 

Partial discharge (PD) technique

Primer for plant operators (from Adwel site) also consult FAQ from IRIS

Many people feel that North american utilities enjoy a lead in using PD techniques to improve pdm of HV generators on-line.
It is thus no wonder that most of the literature on this topic originates from North America.
Furthermore, Internet sites are more prolific there.

There remains an interesting question. Most PD techniques were developed and tested with  success for hydro (air-cooled) generators where insulation aging is more an issue than in hydrogen-cooled turbogenerators. The technique is also marketed for turbogenerators. Does it make economic sense to apply it to turbos? 

The problem of PD focuses upon  aging of the HV insulation in  generator stators. A  book dealing with the issue is Bartnikas.
If some of you know of such similar textbooks, please tell me. Some interesting books in German deal with constructive details of electrical generators (Mashinenbau..)


 

What is Partial Discharge?


Partial discharges are "sparks" involving the flow of electrons and ions when a small volume of gas breaks down. This can occur in voids within the electrical insulation system itself or adjacent to the insulation of high voltage stators. The term partial is used since there is a solid insulation, such as epoxy-mica, in series with the void, which prevents a complete breakdown. The insulation system and the voids can be considered to be a series of capacitances with different dielectric constants. 

Depending on the size of the void, the dielectric constant, and the temperature, the stress on the gas within the void may become high enough for breakdown to occur. In most cases the electric field will not be uniform and this will tend to lower the breakdown voltage.

Partial discharges are often the result of damage caused by other thermal, mechanical, electromagnetic and chemical forces acting on the stator winding. The progressive development of partial discharge activity is the major symptom of insulation deterioration. These discharges also contribute to the aging of the machine's dielectric system by eroding away or
deteriorating the insulation system.

Of course, paritla discharges tend to occur near the line ends where the voltage is highest and not on the side of the neutral point.

                        Ten reasons why you should use Partial Discharge Analysis (PDA) Technology

1. Averts Catastrophic Failures of Stator Insulation Systems
PDA monitoring significantly reduces forced outage time and the possibility of generator or motor damage by giving an early warning of stator insulation system deterioration.
2. Extends the Lifetime of Stator Insulation
PDA readings can identify the early stages of reparable insulation damage (e.g. slot discharge due to erosion of the semi-conductive coating) which, if corrected, will extend the time before a machine failure occurs.
3. Maximizes Production Revenue
Once the PDA permanent couplers are installed, no interruption in service is required to perform PDA monitoring. On-line PDA monitoring of stator insulation condition can allow you to extend the interval between traditional off-line insulation tests and visual inspections.
4. Reduces Capital Costs
PDA monitoring eliminates unnecessary premature rewinding of stator windings.
5. Enables Efficient Maintenance Planning
Using PDA readings, maintenance is based upon the actual condition of a machine's insulation system rather than intervals of time. Also, priorities can be accurately established for the order in which the maintenance or repair of multiple units are performed.
6. Simple and Safe to Perform
PDA monitoring does not require a large external AC power supply and does not expose the operator to high voltage risk. All the skills necessary to perform PDA monitoring can be quickly learned and carried out by station personnel.
7. Gives Specific Information for Assessing Repairs or Rewinds
PDA monitoring can identify the type of the insulation deterioration (e.g. slot discharge), its possible cause (e.g. loose winding) and its severity (degenerating, stabilizing or improving). Also, single severe sources of deterioration can be identified.
8. Confirms Effectiveness of Repairs
Comparison of PDA readings taken before and after repair or rewedging of the machine allows the user to quickly and easily determine if the repairs were successful.
9. Provides the Most Complete Diagnosis of Stator Insulation Condition
Since PDA readings are obtained with the machine in normal operation, all of the thermal, electrical, mechanical and environmental stresses acting upon the winding are present and their effects incorporated into the readings.
10. Non - destructive - Does Not Damage the Winding
PDA monitoring is a non-destructive test. It is performed with only the normal thermal, electrical and mechanical stresses which act upon the winding. This prevents any part of the stator insulation from being subjected to abnormally high stresses, such as the neutral end of the winding would experience in DC Hi Pot testing. 
 

                Partial discharges occur in various places in a rotating machine

1. Partial discharge may exist within the main groundwall insulation as a result of delamination or voids caused by missing or incompletely cured bonding material.
2. Partial discharge may exist at the copper strand/groundwall insulation interface as a result of poor bonding.
3. Partial discharges may exist in the endwinding region where adjacent coils can have full phase to phase voltage between them. Partial discharge may also exist in the endwinding region as a result of tracking (meaning of this?).
4. Partial discharge may be present where the coil exits the slot. Because the outside of the coil in the end winding is not grounded, it acts like a capacitor with one side not connected. Therefore, it will try to assume the same potential as the copper. A sharp change in potential will exist along the surface of the coil between the portion grounded to the stator core and the ungrounded portion. Stress  (semiconductive?) grading paints eliminate high potential differences which would cause partial discharge.
5. Within the slot, the coil surface is grounded to the stator core with a coating of low-resistance conducting paint or tape. Partial discharge can occur within the slot when contact is lost between the conducting surface on the coil and the core. This is known as slot discharge. Slot discharge can seriously burn the surface of the coils and the slot fillers.

In addition to the electrical effects, partial discharge generates ozone (in air-cooled (hydro) generator mostly?). Through a series of reactions, nitrogen based acids are formed at dew point from NOx. These acids will also attack organic insulations, accelerating the deterioration. In some cases, severe partial discharges could ignite an explosion. Again what about hydrogen-cooled turbogenerators? Does some ingress of foreign particles or gas warrant the use of PD there?

Other manufacturers of PD equipment

IRIS (Canada)

IRIS was the pioneer of PD techniques from their early experience with Ontario Hydro. Consult the FAQ section of their site for clear-cut comments. Probably oneof the best sources of info on PD.

Eleco Dowding primer:

Other manufacturers also market PD system like Eleco Dowding. Here is their primer in the pdf below. They use Rogowski coils as couplers. It may be fine for motors , but some calim these coils may not have as wide a bandwidth as 80 pF coupling capacitances  used by other PD suppliers.

Power Diagnostix Systems

A German company also offers PD equipment: Power Diagnostix systems GmbH in Aachen. It also has a branch in Switzerland.

Further readings

Partial discharge documents (one can access protected pdf files by mailing me)

tip up (off-line)

Document from  www.doble.com
The power factor of an insulation is a measure of its dielectric power loss, and is not a measure of its dielectric voltage strength. The power-factor
method is used primarily for testing the insulation condition of high-voltage bushings, cables, transformer windings, and transformer oil.
It is sort of a very simple, traditional and not so sensitive way to perform off-line some sort of global partial discharge test. This test is rarely performed in the Belgian grid, as far as we know.

 

Generator thermal conditioning

with chemical tags

(  from http://www.eone.com/detection/reprints )

The idea is to coat  the windings with trace elements that differ from one lacation to the other. For example, the trace differ for the coating on end coils and in the slots. By analyzing cooling gas with a precision chromatograph, one can identify when and where hot spots may indicate a degradation of the insulation.
Some manufacturers and plant operators express reserves because they fear the technique might cause some unexpected problems.

The following pdf are obtained from the site of Environment/One and explain the basics of the technique. It also addresses the case of air-cooled generators which exhibit some special features. Note that air-cooled generators are more sensitive to discharges than hydrogen-cooled because ozonecan be locally generated and degrade insulation faster.

classical with RTD Thermocouples or optical fibers

Core temperatures are monitored by arrays of temprature sensors.
Thermocouples are lodged in the core to monitor local heating. RTD perform the same task with more accuracy since they produce more robust signals less sensitive to elm noise (not quite see below).
But RTD stands no chance to substitude to partial discharge. IRIS tells you why.
Recently fiber-based techniques to measure temperature transients are emerging. They are insensitive to elm noise and presumably have an ultra-wide bandwidth response not impaired by wiring. If you have interesting links or prints on this topic, please mail me

Another interesting question:
Suppose that one has some severe partial discharge in stator bars either slot discharge (sort of corona) or internal due to delamination. to it is associated some thermal energy. If these occur on end windings, it is perhaps to much to ask.
Is there a way, before it is too late, to correlate transient temperature measurements (supposing one has such fast responding temperature sensors) with devices specifically monitoring partial discharges? That is always a better idea to resort to two radically different types of measurements to assess the same damage. This reduces the rate of false alarms. If you have some ideas on this,  please mail me
 


 
 

Vibrations

Rotor thermal unbalance and shaft vibrations

Uneven thermal fields developing on the generator stators cause thermal unbalances that often give rise to shaft and bearing vibrations.
Such a situation may occur for different reasons:
  • short-turns in the field winding (explained below in some detail although it is not the main cause for shaft and bearing vibrations)
  • rotor cooling ducts whose cross-sections get partially reduced due to a poor thermal expansion of rotor windings.
  • impeded differential expansion of rotor bars in their slots (severest cause giving rise to a combined thermal/mechanical bow).
As a result, the generator ratings (especially reactive power) may get reduced. Although this is far less a serious problem than a faulty stator insulation, it may cause some operating headaches. 
One way to select which of the above causes may produce vibrations is to observe how vibrations react to a step in the field current. 
Time constants related to rotor core and winding heating markedly differ and correspond to variations of vibrations in opposite phases. They will show up in case of a poor differential expansion winding/slot.We had a very severe experience of this kind long ago in Belgium.
Only one time constant will appear in the vibration response for the other two cases, except in machines with multiple pairs of poles. As explained below with the gentech results, a field shortturn may awaken a magnetic unbalance, perhaps not so severe as in induction motors characterized by much smaller air gaps than synchronous generators. The vibration transient is then related to the transient time constant (I believe T'd) of the machine. This is very short compared to thermal time constants.
Anyone is welcome to comment on this...
Manufacturers perform off-line tests to detect field short-turns by reflectometers. 
Other on-line methods are marketed to identify short turns like GeneratorTech.

Gentech system

Gentech html pages describe how they can spot short turns in the field windings by a careful analysis of radial magnetic field in the air gap. To this purpose they place a magnetic probe at the top of a stator slot. They analyze the pattern of mag field by sweeping various combinations of active and reactive power and thus varying the internal angle. This ensures  that each rotor slot is aligned with a minimum mag field to reduce the variation of air gap field due to the passage of rotor teeth in front of the sensor and thus better detect a failing turn . Intricate but quite interesting for data processing aspects.
One can find  interesting information like in the following pdf (less the quite impressive animation (!) of their web pages). Since you are on this academic (slow at times) server, you can download some of Gentech pages (they get the full credit for them)
  • soon links here...
This unique technique may not represent the mainstream of generator pdm. One worries a lot more about other causes of failires.

Mechanical unbalances due to shorted field turns

Gentech has a few interesting commentson this problem. They are contained in their page set up here in pdf  format for your convenience.

It is known that shaft (and bearing) vibrations often react to a field current step with time constants related to bar and rotor core heating. Several factors may contribute to this behavior (see above). Let us focus on the contribuion of short turns.

For a bipolar generator field shortturns may or may not cause a mechanical unbalance.


Modified from source in Gentech site
Let a short turn occur in slot 1 (shortturn #1). Then the current takes a short cut by jumping over one turn from one layer to the next. In the process the corresponding end turn in thick blue  is traversed by less current.. 
Since the number of turns decreases, the field current is bound to increase to maintain the same phase voltage at the generator terminals. If the number of short turns increases, then it may possible to monitor this as long as the excitation is static with slip rings. It is less easy with brushless rotating exciters 
Both slots 1 of pole B (leading and lagging) are traversed by less layers of field current than the slots 1 of pole A (leading and lagging).
Due to the unblalnce of RI2 losses in these slots,  a thermal bow appears whose phase is aligned with pole A.
If a short turns occurs near the quadrature axis (short turn #2 e.g.) then no sizable thermal bow shows up. 
Short turns may or may not be spotted with vibration measurements for bipolar generators. The thermal unbalance, if any, caused by them is aligned with the poles. The  phase of vibrations depends on many other factors like the proximity of a critical speed.

As discussed in the gentech document, short turns on generators with muliple pairs of poles should in principle generate thermal bows. In addition, they genenrate magnetic unbalance that can be identified by their shorter response time to field current steps.


 

Core vibrations

Core vibrations may indicate some abnormal conditions like poor torquing of the core. Unfortunately, it is also subject to numerous other sources that extend to the vibroacoustic domain (slot harmonics) and even in the ultra sound domain.
Monitoring them may be easy, but the interpretation may require a lot more expertise to assess core condition.
Without asking you for too much details and proofs, could you mail me whether you know about links where I could some technical information how core monitoring can get implemented through vibration , acoustics or ultrasound analyses.

 

End winding vibrations

Poor bracing of endcoils (endwindings) may lead to natural frequencies close to the double the power network frequencies and its higher harmonics. This may in turn lead to excessive vibrations of the end coils and damages to the insulation of stator windings thrugh delamination and/or corona (is that right?).
In the past, insulation material could age so that bracing characteristics caused a shift in the natural frequencies. If this brings natural frequencies close to those of the forcing terms (electrodynamic forces due to phase currents), one may be in for bad surprises. Does anyone wish to comment or elaborate on this?

In order to measure vibration of end coils in the presence of a strong electromagnetic environment, some operator have bought optical accelerometers. Such sensors can be found at
http://www.cybernet.be/gatdbalteau/optical.htm  with pictures and basic principles and specs.

Contact the Belgian representative for this type of sensors to know more about  them and where they are used. I heard that quite a few utilities (both european and north american) have bought some.
Of course when it comes to place such sensors, one may wonder on which end coil to locate them unless it is proven that all of them have a generic problem. This may not be the business of pdm as envisioned by utilities but rather a problem for the manufacturer exprimenting a new design. This is a question for debate.


 

Core delamination and insulation breakdown

This section is somewhat related to the purpose of measuring the core vibrations. What one basically wants to check is whether the insulation of stator core lamination is endangered (vibration can spot it?) or already damaged.
A system is marketed by ADWEL to perform this monitoring tasks: EL CID. Visit their site. You will find the next pictures and comments.
EL CID from ADWEL
Digital EL CID
An advanced and cost effective method of detecting insulation breakdown between the stator core laminations in large motors, turbo and hydro generators
  • Provides consistent and reliable information on the condition of the stator core in less time and at a lower cost than any other technique
  • Direct output to PC's for fast analysis and interpretation
  • Fully digital processing - Phase and quadrature information gathered simultaneously
Analog EL CID
  • Data Interface System (DIS) available
  • Analog metering system

  • Pinpoint accuracy of fault location

Other manufacturers yet to identify, if any....



 

Wedging of stator bars

Motion of stator bars may be restricted in several ways: hard, semi-flexible (ABB?), flexible with ripple springs (in Belgium).
Question to the attention of manufacturers of the different wedging types:
  • How does hard wedging accommodate differential core/bar expansions?
  • May the semiconductive coating at the surface of the insulation (if any) get damaged?
  • With flexible ripple springs, how to make sure that wedging does not degrade in time?

  • Some experience with poor aging of ripple springs brings this question out for debate.

Ways to detect wedging tightness

Adwel System:  WTD Model 501  and RIV for robot inspections of various types.

WTD model 501


 






















 RIV picture 
All pictures from Adwel site

Wedge Tightness Detector WTD 501 for consistent and reliable assessment of wedge tightness through electronic tapping.

RIV
The left-hand side picture corresponds to its implementation in hydrogenerators and requires to remove the rotor. It shows how core control is automated with RIV (Robot Inspection Vehicle).
Low profile robotic delivery system for efficient stator core testing. The magnetically supported RIV was developed primarily to provide automatic scanning of the sensing head of EL CID (see above paragraph for testing core lamination) test equipment throughout the stator core of a generator or large motor in order to test the integrity of the stator lamination. The equipment can be adapted to carry other lightweight transducers such as mini-cameras and wedge tightness probes. ADWEL's RIV has been used successfully in EL CID measurements by leading OEM's and power utilities on large generators of varying designs. The constant scanning speed provides more consistent test traces than those made with hand-held sensing heads.


 

Laborelec system

Controls the resiliency of ripple springs. Pictures coming soon.
 

Traditional off-line winding insulation tests

Stator phase winding

A current test for winding insulation is using the Megger device whose principles are explained  in this pdf from AVO with the definition of the PI (polarization index).
Another paper by AVO introduces the technique of dielectric discharge (DD) which can be used along the technique of polarization index. Useful primer 

Field Shortturns

Reflectometers.

 
 
 

And now what would be a good candidate amongst generators of a fleet to initialize a pdm strategy on it?

When implementing a pdm strategy on hv generators, on,e undoubtedly would wish to select a "good" candidate, meaning a machine where pdm could reasonably present a fair chance of a return on investment. I submit a few criteria for a such a selection.

The table below may be a guide how to select a generator unit to initiate the pdm on it.

Past stable operation

A unit might be weakened  by the number of hours logged at different loads. Pdm tools already installed may signal such units. One may then extend the array of pdm modules if the unit warranst it. The number of starts of the generator may also have shortened its useful lifespan  just like unbalance loading.
 

Transients

Short-circuits (symmetric or others), mis-synchronization (how far it is allowed according to the specs of the manufacturers), rapid fault clearing (causing torsion transients), etc. are as many events prione to shorten the useful lifespan of a machnine.

Past outages

The plant operator can certainly remember how many times a given generator underwnet unplanned outages and heavy repairs. This is a good and evident way to identify troublemakers.

Chemical pollution

May contribute to early aging evidently. In most power stations, the role of these environmental factors are usually kept at bay. In cogeneration unit in refineries and other more polluted plants, this is a factor to consider.

Feedback from maintenance

Not forget what the maintenance crew has observed when repairing generators. Machines which have not been maintained for a long period might also hide serious problems. Thus longer maintenance intervals may point to troublemakers. Of course, a machine may also suffer from too closely spaced overhauls since these can result in human errors.

Spare parts

Obviously unique machines in a fleet may not be provided with all spare parts. PDM then becomes quite an interesting alternative.

Anticipated duty of generators

It is no bright idea to implement pdm of generators that might be declassified in the near future. Better focus on units that are thought to operate for many years to come.

Design features of the generators

Some design of generators are thought more sentivitive than others to early aging.
Some people suggested  me that insulating material using polyester insted of epoxy as bonding were more sensitive to failures. Is that right? In that case, better implement a PD strategy for machine with mica-polyester (polyester is much more sensitive to oxidation) than with mica-epoxy. Mica-polyester still exists in some older generators.
In Belgium, wedging the stator bar is flexible with ripple springs. For some reason or another, these ripple springs sometimes lose their resiliency and wedging degrades. This may cause the loss of semiconductive coatings and subsequent corona (slot discharge). The question is: is it better to monitor wedging at overhauls or monitor its consequences in terms of PDs? Is there a way to monitor wedging on-line? If someone has an idea on the matter, please mail me.
Corona with air-cooled generators produces ozone which in turn acts as an aggravating factor to impait the insulating material.
Endwinding (end coil) bracing should not age to bring resonance frequencies closer to the grid frequency and its higher harmonics. In contrast to metallic support in classical mechanical engineering (loose and interference fits are well documented in the ME literature), one deals here with insulating support whose aging may cause such shifts of natural frequencies of the endwindings (fits with organic material are far less known). This may be bad news...
Also check the specs when ordering a generator or consult them: some machines may not sustain transients more severe than three-phase short-circuits.

Of course, better design is always better than better monitoring. Manufacturers will all agree on this evidence.

Air-cooled vs. hydrogen-cooled generators.

Key to stator insulation aging.

Rotational speed

Faster rotating turbogenerators are more sensitive to thermal bows due to uneven temprature distributions in the rotor. Shaft and bearing vibrations are far less acute in slow-rotating hydrogenerators.

Table to select the most promising generators for starting a pdm

Fill the table below for the generators of your fleet and then select the best candidate... Table