Dear Sirs,
Two question, if it's possible.
1. Motor can be supplied or directly through network (380 V), or through the frequency
regulator.

We have not enough experience in latter case, but if we apply directly MCSA approach for this case, the number of faults is relatively low.

Do you have data for the !!same motor (better obtained in one day), which is supplied
a) Directly from power supply 380 V
b) Through PWM regulator

It is very interesting compare results (or model them by
calculations) for those cases.

2. Often we asked to diagnose motor, which during 15 sec 2 times runs, 2
times stops, or diagnose motor during it start (time-dependent process)

What is your opinion about wavelet usage for this purpose?

--
Best regards,
Victor Petoukhov http://www.motor-diag.com

Hello Victor,
Welcome on the board! Your post is very intriguing. And, by the way, it is very exotic also. But I would say it will not be exotic for too long in this shrinking world.
But let me express my opinion, if I understand correctly. If you find a fault in a motor running from a sine wave 380 Volt, 50 Hz supply, you should find the same fault even if you run the motor from a variable frequency drive running at 50 Hz.
I consider the today’s variable frequency drives sophisticated enough to simulate the harmonic sources closely enough.
If there is disagreement, it is the MCSA that is not sophisticated enough.
I have never compared the current analysis in those 2 cases. And I have to admit that the wavelet usage is way over my head. I bet we all could learn a lot from you. Is there any translation to the attachment?
jank

Dear Jank, thank you very much.

No, I want to compare EXPERIMETAL results for the same motor for this 2 cases (if such exist, of course).
It is clear, that motor faults are the same, but the results of MCSA (or Instantaneous Power, or Park's vector approach) diagnostic may be different).

About wavelets. Please, see http://www.motor-diag.com/index_eng.html, news from 26.11.2007.

And what attachment do you mean?

Victor

By the attachment I mean the link http://www.motor-diag.com/index_eng.html, . It leads me to a site that I have difficulty to understand. With my limited knowledge of the Russian language I could understand for example that current spectrum of a motor (dvigatel) changes over 5 year period (???). I cannot understand why it would be the case (if I understood correctly!).
jank

Dear Jank, thank you.

quotation
By the attachment I mean the link http://www.motor-diag.com/index_eng.html, .
It leads me to a site that I have difficulty to understand.
quotation

Ok, please try URL http://www.motor-diag.com/index_eng.html (the SAME
URL, but WITHOUT COMMA at the end. It works, I checked it right now.
News from 26.11.2007 says: "Now you can download your data (*.txt-format) on Motor Diagnostic and Power Quality on www.rapidshare.com and send e-mail us on a_and_a@mail.ru.
We shall process those data and send you the results of processing with our comments.
First processing of your data is free."

quotation
With my limited knowledge of the Russian language I could understand for example
that current spectrum of a motor (dvigatel) changes over 5 year period (???).
I cannot understand why it would be the case (if I understood correctly!).
quotation

Not, not correctly, of course. 5 year period- from where is it?

I asked in my post:
quotation
2. Often we asked to diagnose motor, which during 15 sec 2 times runs, 2
times stops, or diagnose motor during it start (time-dependent process)

What is your opinion about wavelet usage for this purpose?
quotation

Where is in my post words "5 year period" ???

If your question is about materials on our site, please use
sign "Letter" on page http://www.motor-diag.com/index_eng.html to contact us.

Victor

I have followed your instructions and I have indeed got to your site in English. And I have read a lot with great interest.
There is indeed a current spectrum comparison of a new motor (dvigatel) in part titled “Primery Spektrogramm” in the part: http://www.motor-diag.com/example.html . The left column has a title “Novij dvigatel” and the right column is called AHU-8 (porabotavscij). There is a lot of differences in those 2 spectra! Anyway, that is what I was talking about in my last post.
Let me discuss the first item in your list of possible motor faults you are able to detect: It is the stator turn to turn short. There are lots of vendors in North America that claim the same ability, mostly with the method of “Low voltage testing”.
I have been in the motor repair business in past about 25 years. During those 25 years there was not one customer that would bring the motor to the shop saying: “Listen Jan, I have this problem with my motor. There is a stator turn to turn short, the motor does not perform as it should be, will you rewind it?” Not once! Isn’t it a bad luck?
I can list you tons of customers that brought a motor to the shop with a big hole in the winding. And the big hole is the logical result of any turn-to-turn short in any winding. The current in the shorted turn is roughly the same as the locked rotor current. How do you expect to operate part of the winding with current 5 to 7 times larger than the nameplate current? How much time do you have to detect such a fault? And even if you detect it on time (by miracle), what are your options? Do you have any procedure to extend the life of the motor till the next scheduled shut down?
Is there a chance that the current in the shorted turns is limited to a sustainable level?

It is my opinion that the above questions are more pressing that the wavelet usage. At the same time I have been in a situation where there was a need of analyzing rotor condition of a motor in a wood chopper application. The time window for such an analysis was less than 15 seconds and the load was all over the place. I have to admit that I have failed miserably.
jank

Jan, thank you.
But I asked in my posts:

1. What is your opinion about wavelet usage for this purpose?

2. If your question is about materials on our site, please use
sign "Letter" on page http://www.motor-diag.com/index_eng.html to contact us.

I must say, that we do not diagnose motors with
"a big hole in the winding"
It is too late, and fault of motor is obvious.
Please, send us data of such motor BEFORE SC fault, we process those data and only after that we can to discuss those data.

And I do not understand this phrase: "...and the load was all over the
place". Please, say it in another words.

And I appreciate e-mail you directly, not on this forum.


Victor

Hi Victor,
Forgive me my skepticism. But I had to read at least part of your site, because I wanted to make sure that I am not dealing with another vendor selling a miraculous gadget that will tell you not only all maladies the motor suffers from, but also the shoe size of the motor designer. All you have to do is hire a guy from the street, teach him a sequence of few strokes on the computer, and the computer will tell you the rest of the story. I have learned that this scenario indeed does not work.
However your question was quite clear:” What is your opinion about usage of wavelets”.
I will be brutally honest with you. I do not have a clue! On the top of it I do not seem to have the incentive to go through all the trouble to learn it. You are in a much better position. At one time you have decided that the Fourier was not enough. Why don’t you tell me what the breaking point was?
I imagine that using the wavelets you can actually predict that the motor stator turns are destined for disaster (?????)
I have read numerous articles where the experimenter introduced a stator short and then identified the frequencies this short generated in the current spectrum. I imagine you have mastered this art, am I correct?
…the load was all over the place,… means that during the analysis the current varied so much that it rendered the current analysis totally useless. In the end everything ended peacefully. The static test (with the motor disassembled) confirmed, that after 4 years of operation the bar # 16 broke (total of 57 bars). See the attached ppt file.
jank

broken_bar.ppt (61 Kb, 15 downloads) comparison of the condition of the rotor bars after 4 years of operation

Dear Jan,
> Forgive me my skepticism. ..., because I >wanted to make sure that
> I am not dealing with another vendor selling >a miraculous gadget that will tell you not >only all maladies the motor suffers from, but >also the shoe size of the motor designer.
Yes, you are quite right, and we try to speak less (and only about experimentally proofed items), and to do more.

> I have learned that this scenario indeed does >not work.
Yes, this scenario does not work, because there is no miracles.
But guys from any factory want to know: "What is wrong with THIS motor,WHEN it will be out of operation and what do to with THIS motor", not more.

> However your question was quite clear:” What >is your opinion about usage of wavelets”.
> I will be brutally honest with you. I do not >have a clue!
Thank you for the honest reply.

> At one time you have decided that the Fourier >was not enough.
Not we. FFT (sin and cos decomposition) works only for stationary signals(sin and cos are stationary functions). For time-dependent
signals it works only formal and not correct.

> I imagine that using the wavelets you can >actually predict that the motor stator turns >are destined for disaster (?????)
Not exactly, the question is more complicated.

> I have read numerous articles where the >experimenter introduced a stator short and >then identified the frequencies this short generated in the current spectrum. I imagine you have mastered this art, am I correct?
Yes, of course. But we made such experiments ourselves.


Victor

Victor,
I would like to continue the discussion about the stator shorts. There is an interesting paper by Marian Dumitru Negrea from Helsinki University. http://lib.tkk.fi/Diss/2006/isbn9512284774/isbn9512284774.pdf
There are very valuable information on every page, but let’s read, for example, from page 20: “Excessive heating caused by turn-to-turn shorts is the reason why the motors in this condition will almost always fail in a matter of minutes if not seconds”.
This time schedule of the failure is in agreement with my experience. The short period of time between the occurrence of the short and a catastrophic failure makes the task of the short detection extremely hard.
However there is one problem that is even harder to overcome than the rapid progress of the failure: There is no real current to detect!
The experimenters, that determined what harmonics to look for in the shorted winding, usually created and artificial short in the winding. But it has never been a perfect short. They usually inserted a limiting resistor outside the motor to hold the current at sustainable level. Sometimes the resistor was even water-cooled!
I have seen many motors from inside and can confirm, that there is no room for any limiting resistor particularly between turns for the turn-to-turn short. Usually there is a layer of 5 to 6 thousands of an inch of high quality enamel between the wires. If the failure starts in an area of 1 mm square or 1 cm square, the mass of this “resistor” (that is supposed to limit the current) is always only a few thousands of one milligram. This “resistor” cannot absorb any meaningful energy. In other words it cannot be a current limiting resistor as is the experimenter’s water-cooled resistor.
Yet we are constantly hearing about fantastic methods of predicting the turn-to-turn short with methods of low voltage testing. In this simplistic world the current through the fault is first 1 mA, then 20 mA, than 1 A and so on. In other words there is deterioration, then there is more deterioration and than there is even more deterioration. They even double the frequency and pretend it has an important effect!
For anybody who wants to detect an impeding short, the current through the FUTURE short must be detectable. But the current through the future short is not detectable by simple means. The wires are either touching and it is a perfect short. The catastrophic failure is only few seconds away and nobody can stop it because it is buried deeply inside the winding.
Or the wires are not touching metal to metal. In this case there might be some low energy arcing between those two wires once a while, or there is no current through the FUTURE short at all, especially with LOW VOLTAGE. The “once a while” discharge with duration in nanoseconds and negligible energy cannot be detected as a current increase, hence it is NOT detected.
Note: There is a ton of other reasons for current change when changing the frequency. None of them is related to the current through the future short.
However the sporadic discharges can, and eventually will, initiate a high-energy discharge and a “hard” short.
What does it have to do with wavelets?
I have done some quick reading on the wavelets and I am impressed. First of all the wavelets can treat the discontinuities, the Fourier cannot. The turn-to-turn short likely starts as series of low energy discharges (arcing) in the affected area. If the wavelets can catch those discharges, there might be hope for an early detection.
You mention that you conducted some experiments. If those experiments involved a limiting resistor outside motor, it would not be anything new and you would not need wavelets. But you may have gone a little farther. I do not expect you to reveal the results of your hard work. But it would certainly be an oasis in the present low voltage desert.
One more interesting observation from the above link. This serious and comprehensive study does not mention the low voltage testing even once. I wander why?
jank

Jan, thank you very much for very serious reply.

> There is an interesting paper by
> Marian Dumitru Negrea from Helsinki >University. >http://lib.tkk.fi/Diss/2006/isbn9512284774/isbn>9512284774.pdf
Indeed, very interesting and serious work, it must be read "with pencil".

> This time schedule of the failure is in >agreement with my experience.
With our, too.

You are quite and absolutely right, there is no miracles in electrical engineering (I speak about "fantastic methods"). But if you look at our protocols, you will see: "NONSYMMETRY OF PARAMETRS (SC) OF STATOR WINDINGS". Word SC is in brackets is not by chance. Nonsymmetry of parameters MAY LEAD to SC, but may be due to random rewind of motor and so on.
We saw such motors 1.5 years ago, and ONLY PART of them are burned out (they had SC).
The only thing about we say is Nonsymmetry of parameters, and to this circumstance must be paid attention, mot more, but not less.

About wavelets:
You are right, they can be used for early diagnostic, but with wavelets everything is not so simply.
1. When we change type of wavelet, we receive another frequency picture...
2. All diagnostic frequencies for different faults are ONLY for Fourier (Sin and Cos) decomposition, not for wavelets...

> I wander why?
You answered yourself for this question in your reply:
>1. Or the wires are not touching metal to >metal. In this case there might be some low >energy arcing between those
> two wires once a while, or there is no >current through the FUTURE short at all, >especially with LOW VOLTAGE.

>2.The catastrophic failure is only few seconds >away and nobody can stop it because it is >buried deeply inside the winding.

I wonder another thing: why we do not see change of current spectrum values on predicted for SC failure frequencies, even for laboratory
metal SC without any resistor? Another physics of failure?
Victor

I would like to further discussion on Jan's previous post.

While troubleshooting an electrical circuit (not a motor) I checked continuity of a fuse with a Fluke meter and it showed good. Luckily, I checked voltage across the fuse and it showed open. There must have been carbon tracking across the blown fuse. If I had not performed the second check (high voltage check), my troubleshooting would have led me in a stray direction that could have taken a long time to get back to re-checking the fuse.

The analogy is that the fuse was open as far as carrying the required operating current at line voltage, but closed as far as reading it with a voltmeter.

My point is that lab approximation of field condition can never exactly replicate field condition. Likewise, to say a short will be either a perfect short or a future short is not an exact description of field conditions.

In my experience electrical shorts create carbon. In certain conditions, the carbon track can be read by low voltage testing means.

I only want to say that I believe neither side of the argument is ever black and white.

J-

One has to agree with your point that the laboratory conditions do not replicate the field condition exactly. But I do stand by my previous assessment of the SHORT and the FUTURE SHORT. The people that sell you the low voltage testers do not tell you that they can find a short, they are selling you equipment that is (in their view) able to find the future short before it happens. As a result you can get ready and the production is not affected. That is the selling point. When there is a short in the winding the game is over. Period.

quote:

Originally posted by Wally:
In my experience electrical shorts create carbon. In certain conditions, the carbon track can be read by low voltage testing means.

I only want to say that I believe neither side of the argument is ever black and white.

J-

Let me now examine your above statement. For explanations I will use data from the 800 hp motor because it was just recently wound and I have all necessary data. The motor is low voltage (575Volts) but the winding is not random wound, but the form coil type. The turn-to-turn voltage is 575Volts/33(turns in series)= 17.5 Volts. The connection is Delta.
Now imagine that the insulation between 2 wires of one coil deteriorates for some reason (for example due to carbon as you have suggested above) in an area of (let's say)2x2 mm (those faults are always very localized). Imagine that the resistance of the turn-to-turn insulation that was originally virtually infinite in that spot drops to a mere 1000 Ohms. Since there is 17.5 volts across this “resistor” current will start to flow I=17.5/1000= 17.5 mAmps. This “resistor” will have to dissipate V^2/R =17.5^2/1000= 0.3 watts. No big deal. The heat will dissipate to the surrounding material (mainly copper), and not very much will be happening. However the spot will be a week one.
Now what happens when the resistance drops to 10 ohms? This tiny resistor now has to dissipate 17.5^2/10 = 30 watts! There is absolutely no way that this “resistor” can last more than few seconds. If you want to see what’s going to happen, get a ¼ or ½ watt 10 ohm resistor and apply 17.5 volts to it. There is just a brief spark, little smoke and the resistor is gone. Something similar will happen with your carbon track between the turns. Little spark that ignites an arc between the wires and hard short will follow. End of the story.
Let’s now examine if the low voltage testing can catch this “future” short. A guy hooks up his tester of whooping 9 Volts across the terminals. The turn-to-turn voltage will be 9/33 = 0.27 volt.
Since the motor is in locked rotor condition, the stator winding will now “see” 2 secondary windings. One is the locked rotor and another is our almost shorted stator turn. The voltage induced into the shorted turn will be roughly 1/32 * 9= 0.28 Volts. (The number 32 is 33 turns in series minus 1shorted turn). Neglecting the reactance (which is extremely pessimistic), the current will be 0.28volts/10ohms=0.028 Amps. This will show as 0.028/32 = 0.00088 Amp on the primary due to transformer action (and on the tester). There are other influences (such as parallel branches) that would diminish that number even further.
Yet we also know that the locked rotor is something like 6 times the nameplate current of roughly 800 Amps. Hence the impedance is 575Volts/(6*800Amps)=0.12 Ohm, in single-phase mode 0.076/0.866=0.138 Ohm. So the other secondary ( the rotor) will ask for 9Volts/0.138Ohms = 65.2 Amp. Somebody may protest because the tester uses high frequency such as 600 Hz. Yes, that is true, but it will limit currents in both secondaries, they both have reactances that grow with frequency. Changing frequency is virtually irrelevant. So our low voltage tester has a task equivalent to distinguishing between current of 65.2 Amps and 65.20088 Amps. Any reasonable technician who has ever held an ammeter in his hand knows that it is impossible and the same applies to the low voltage tester. For a different motor size it may translate in to distinguishing between 6.520000 Amps and 6.520088 Amps. The same impossibility. Of course for higher resistance than 10 Ohms the chance of discovering the future short is even less likely.
In other words the low voltage tester does not have even a remote chance to see the future short that will inevitably end as a failure of the winding. Yet some people see them as a panacea of the motor testing. In many cases those are the people from maintenance departments of large companies that can afford any toy under the sun. They replace the knowledge with enthusiasm. Isn’t it neat: You hook up your computerized tester to the input leads and in few keystrokes and few seconds later you know everything.
NOT.
I have tried to estimate the conditions by calmly looking at the facts. Somebody may not agree with my calculations. I would like to hear from him/her. But from the above one can conclude that the “low voltage means” are NOT able to read the future short (that begins for example as carbon tracking) and that there ARE things just black and white.
jank

Jank,
The fuse comparison is not the same way a motor winding short occurs; I believe that to be true. It wasn't meant to be a direct comparison. I also don’t have a lot of faith that the low voltage testers can find a potential short or weak insulation. I just have a hard time putting anything to do with electrical problems in black and white terms. Megger readings can be different at different humidity levels. A short will track at some humidity/distance/dirt combinations but vary one or two of those factors and it may not, or it may start to track and then stop as it creates its own new micro environmental conditions at the site of the short.

Let me think about the voltage comparison you made between operating voltage and 9v tester voltage a moment though.

I like the transformer analogy. Let’s take say, a 480/120v high current transformer, let’s make it a 500A single phase transformer. Will it also function properly as a 9/2.2v transformer? Or will the large primary conductor act more as a conductor than a primary winding at 9v? Will the magnetic force generated by the primary be able to transfer through the core to the secondary at 9v? I don’t know the answer. But if not, the primary would act more like a conductor, is that right? If so, I don’t know how that would apply to the formulas involved at 9v. Your 9v comparison must also assume an infinite 9v supply to generate the numbers stated, as would be the case using the utility.

There is more to this than I know the answer to, but its fun to discuss theory of it.

J-

If you did not asked I would have never figured out that the linearity of a transformer may be in doubt. Just a quick review: I am attaching a ppt file where I measured the output voltage of a 500 VA, 2400Volt to 120 Volt transformer. The highest voltage on the 2400 volt side was 30 Volts. The corresponding low voltage was 1592 milliVolts on the low side. The lowest voltage on the high side was 0.187 Volts and output was 10.4 milliVolts. Both from the linear graph and from the log-log graph it can be seen that the transformer is incredibly linear to the lowest possible voltages.
Hi voltage was measured by Fluke 77, low voltage by Fluke 87.
jank

linear_transformer.ppt (52 Kb, 16 downloads)

That is very linear isn't it. Nice report, thanks for trying the test. I wonder what would happen if the test were more scaled to yesterday's question.

1) 60 to 150 kVA 480/120 single phase xformer to simulate motor (we replaced one recently to upgrade the system- probably gone now, I'll check).
2) Variac preset to 9v (we have a 10A rated variac), connect by rated switch to primary to simulate limited supply test instrument (probably should be fused to protect variac from overload)

A large transformer takes a pretty good inrush to magnetize the windings at rated voltage. I just wonder, if with a limited 9v supply, that it might just overload the supply before the windings magnetized.

A power resistor could be connected to the secondary to simuate a stationary rotor (I'd consider the rotor a load until it came up to speed). I have several 47 ohm 200 watt power resistors in a drawer collecting dust. I'm guessing that just powering up the windings without load might just overload the variac though.

That's more what I was thinking. I don't think that I can put this test together in a short amount of time.

Back to the test instrument. I believe Baker Instruments uses a high voltage capacator to discharge into the motor windings, then measures decay. Could the low voltage test instrument measure in a similar way (I know nothing of how they work). I'm assuming the test instrument cannot achieve the motive force neccessary to magnetize the windings without voltage drop.

Jank, are you anywhere near the Tar Sands sight? I was out there for a visit many years ago and got to tour the sight with someone from the inside. It interested me because I was at Unocal's Oil Shale site in Colorado. It is a similar, but the process of removing oil from shale is harder on equipment.

J-

Let's make that a 3 phase transformer. They are easier to come by and will better simulate a motor.

I looked at the All Test Pro online. I don't think that unit could supply 65 amps to the secondary, do you? They must be analyzing the winding some other way than a straight measure of amps as your calculations suggest. The impedance on a large motor or transformer is just too low for a limited power supply to keep the voltage up. I don't know what else to think.

We are awfully far from wavelets. My sincere apology Victor!

Wally, the transformer is a poor simulation of a motor. If you want to experiment, there is always a motor somewhere. Any low voltage tester sees the motor in locked condition, it means the reactance is overwhelmingly larger (unlike your resistors on the secondary of a transformer - the load would be overwhelmingly resistive). Your 150 kVA transformer is about 200 hp. The tester would see something like 0.025Ohm resistive load and 0.25Ohm reactance at 60 Hz. So if it were 3 phase motor it would need about 9Volts/0.025Ohms=36 Amps. In single-phase mode of measurement: 36* 0.866=31 amps from the 9 volts tester. Since it is impossible with a little battery powered toy, they increase the frequency to 600Hz or 1200 Hz. It limits the current by the factor of 10 respectively 20 by increasing the reactance. All those manipulations decrease the flux density in the airgap dramatically. If the flux density in normal operation on your 480 volt motor is 0.8 Tesla (8000 Gauss), the flux density with the low voltage tester at 9 Volt and 1200Hz is 0.8*(9/480)* (60/1200)=0.00075 Tesla (or 7.5 Gauss). The accuracy of the results is similar as if you measured a bearing with a tape measure and even worst.
Your mention of Baker: I believe you are talking about surge testing. A classical test and the only test for turn-to-turn insulation. Excellent and reliable test. The low voltage testers do not use this method. They generate low voltage sine wave with frequencies substantially higher than 60 Hz. How close to the ideal sine wave I do not know.
The Tar Sands are just few hundred miles north. I go there frequently.
jank

It would seem that what Jank is describing in terms of difficulty of using low-voltage testers to do winding condition analysis has been my exact recent experience during a low-voltage tester demonstration.

So as not to hi-jack Victor's thread, I have started a new thread.