Capacitor safety vent audible sound

[morwind]

Hi I am a degree student in the UK, I am looking at designing an early detection sensor for the detection of failing capacitors in variable speed drives in lift shafts, during operation. Albeit the sensor could be used in a wide range of remote unmanned applications (substations etc).

So the entire basis of my sensor is on the theory that when capacitors vent due to electrolyte breakdown they start to produce hydrogen gases inside the can which is then vented gradually through the safety vent over a period of time. Note: not instantaneously like all the "blowing up capacitors!!!" videos around the internet.
I am talking about during routine operation, not reverse biasing, over voltage etc....

My sensor would detect the audible noise emitted during the venting process and when detected an event would occur such as: disconnect vsd supply voltage. Preventing further damage to the circuitry.

TLDR: Am I right in thinking that there is a detectable/measurable audible noise when the aluminium electrolytics vent? Which increases gradually as the capacitor vents.

[Ian.M]

Your assumptions are incorrect.  There may be an initial 'pop' of some sort - either the bung to can seal failing or the can splitting at the stamped vent lines, but once that's happened there is unlikely to be sustained noise from the failing capacitor detectable over the usual background noise level in a plant room, let alone a lift shaft, and the pop is going to be near impossible to distinguish from random background noises.  Increased ripple on the DC bus may result in increased noise from other components but that's going to be extremely model specific.

You might get somewhere with a chemical sensor to detect vapors of typical capacitor electrolytes ...

It may be better to tackle this at the VFD design level.  It would be reasonably simple to detect excessive DC bus ripple and depending on its severity, either indicate a fault and enter a degraded performance 'limp home' mode or trip out.

[wraper]

If capacitor fails gradually, there will be barely any noise when vent ruptures, if it ever ruptures at all (in most cases it does not). Noticeable sound will happen only if capacitor is put under extreme stress which will create rapid pressure increase within the capacitor.

My sensor would detect the audible noise emitted during the venting process and when detected an event would occur such as: disconnect vsd supply voltage. Preventing further damage to the circuitry.

If your sensor will be so sensitive to have ability to detect anything at all, there will be orders of magnitude more faults created by false positives, than actual failures detected. It's a tech which has no use in the real world.

[ogden]

Many most electrolytic capacitors don't die with bang. They gradually "dry out", lose capacity until circuit can't tolerate ripple anymore.

[Benta]

I'd think that a more viable way of keeping an eye on electrolytic capacitor health would be monitoring ESR.
Just an idea.

[The Soulman]

I'd think that a more viable way of keeping an eye on electrolytic capacitor health would be monitoring ESR.
Just an idea.

Yes, simply monitor the ripple voltage in normal use.

[Siwastaja]

Monitor DC link voltage ripple knowing the current ripple (the inverter generates and measures the current ripple). Then you can accurately calculate ESR = dV/dI, and react to out-of-spec value.

Do not stop the elevator on increased ESR. Instead, report the failure: needs maintenance.

Don't make the limit too close of the original, expected ESR. Choose the threshold so that there are no false positives, to get a report only when the unit is actually close to failure. Otherwise technicians start to ignore the warnings soon.

Finally, design properly with good capacitor brands and derating so that expected lifetime will be well over 10 years.

Additionally, you can measure the temperature of the capacitors.

[tom66]

As others have said, detecting the capacitor vent sound may well be impractical.

However, during the venting process, large amounts of hydrogen gas are produced - perhaps a hydrogen  sensor could be used?  A little hint: most natural gas sensors based on tin-dioxide elements are also cross-sensitive to hydrogen gas.  It is theoretically possible to use these devices in a hydrogen sensor application.  (I have used Nissha SB-12A before.)  You could also use a gas sensor tuned specifically for hydrogen gas.  These sensors are not too hard to use, although they do require calibration against a reference gas. If your goal is to detect change though you could do it with minimal calibration.

[wraper]

Measuring ripple is pretty much of little use. It will detect only one very narrow type of failure when there are many more others. Not to say that due to overengineering, reliability likely will be lower than without such monitoring. It's better to invest the resources into actual increase of reliability, say using more capacitors to decrease current ripple per capacitor or using better parts. I don't say than none of parameter monitoring is needed, just probably not of this type.

[Siwastaja]

Yes, I think if capacitor "state-of-health" measurement is needed, then ESR monitoring is the simplest, most reliable, and most useful way to do it because being a motor drive, the known current ramp already exists, only thing left is measure voltage ramp. The problem is this only shows the combined parallel ESR and can't detect any single cap starting to fail. For that, you would need to add per-capacitor temperature sensing which would still be much easier than the crazy sound-based idea.

All in all, I don't think any of this is good use of engineering resources. Instead, learn about the failure modes and how to prevent them.

[ogden]

Yes, I think if capacitor "state-of-health" measurement is needed, then ESR monitoring is the simplest, most reliable, and most useful way to do it because being a motor drive, the known current ramp already exists, only thing left is measure voltage ramp.

This is other way to say "measure ripple voltage", at given load/current. Power bus ripple peak detector does not sound like bad idea for hi-end devices, nor it would be complicated circuit. One needs AC decoupling, rectifier diode, lowpass RC filter and comparator or ADC and means of reporting problem.

Idea for OP: you can detect bulging/ruptured capacitor by gluing thin/brittle wire over it's rupture area and check continuity.

[wraper]

This is other way to say "measure ripple voltage", at given load/current. Power bus ripple peak detector does not sound like bad idea for hi-end devices, nor it would be complicated circuit. One needs AC decoupling, rectifier diode, lowpass RC filter and comparator or ADC and means of reporting problem.

But you likely to have several power buses and capacitors in places hard to measure, say in primary side of SMPS. In many places dead capacitor won't increase any ripple to measure at all. Say would you bother to measure ripple on Vcc/Vin of PWM controller IC on primary side of PSU? Yet capacitor sitting there is among parts which fail the most often.

[Ian.M]

Use a crappy small electrolytic there in a Muntzed PSU design, put it right next to the chopper heatsink, and it will certainly be a leading cause of failure.  Small can electrolytics are notorious for not tolerating high temperatures and ripple currents well due to their small surface area for heat dissipation and once they start to degrade, due to their extremely low reserve of electrolyte in their tiny volume, rapidly fail totally.   Use an appropriately rated well chosen solid dielectric capacitor and you can virtually eliminate failures due to excessive ripple at the primary side SMPS controller power pin.

[ogden]

But you likely to have several power buses and capacitors in places hard to measure, say in primary side of SMPS. In many places dead capacitor won't increase any ripple to measure at all. Say would you bother to measure ripple on Vcc/Vin of PWM controller IC on primary side of PSU? Yet capacitor sitting there is among parts which fail the most often.

You always can find circumstances where implementing safety device fault detector is impractical. If somebody have illusion that capacitor failure detector shall be built-into every device where electrolytic capacitors are present then he is mistaken indeed Yes I agree with you that proper design and proper choice of components shall be paramount. If  those two are not enough only then next level of safety reliability measures shall be engineered.

[Siwastaja]

This is other way to say "measure ripple voltage", at given load/current. Power bus ripple peak detector does not sound like bad idea for hi-end devices, nor it would be complicated circuit. One needs AC decoupling, rectifier diode, lowpass RC filter and comparator or ADC and means of reporting problem.

Actually any properly designed motor drive already measures the DC link voltage for various reasons: control algorithm feedforward, to stop regenerating and let the motor freewheel if DC link voltage rises above maximum...

So really the only thing to do is to make sure the existing measurement circuit has enough bandwidth - maybe by reducing the amount of RC filtration, and increasing sample rate in software. If you sample twice per output stage PWM cycle, you have enough data already to calculate ESR on the fly: you know all dV, dI and dt.

This may be zero hardware cost and not very complex on the software side, too. The biggest hurdle is to decide how to report and handle the error. And, avoiding false positives. For example, cold capacitors have higher ESR, but this isn't a serious problem because the capacitors will self-heat so the ESR lowers back to normal range.

wraper has a point though, even if you monitor the trivial motor controller DC link, the big picture will have dozens of power supplies with dozens of DC links that are not trivial to monitor. By Murphy's law, everything else fails except your fancy monitored motor drive DC link. This is unsurprising: often the small elcap in an auxiliary supply fails first, maybe because the designer didn't think that supply is critical.

[ogden]

wraper has a point though, even if you monitor the trivial motor controller DC link, the big picture will have dozens of power supplies with dozens of DC links that are not trivial to monitor.

Most modern electronics devices has so many components which could fail that monitoring health of everything is impractical. Isn't that bloody obvious? We suposedly talk about devices which are required to meet some reliability and/or safety standards. We supposedly do not talk about generic SMPS of somebody's generic desktop PC. Think of hi-rel & hi-cost medical or industrical devices.

[Siwastaja]

Yes, but even that motor inverter may have a few auxiliary supplies in it (maybe three for isolated high side gate drive, one for logic board, and so on), and if you think about other industrial modules in the same system, all critical for the complete system to function, the count quickly goes to dozens. These small supplies are relatively harder to monitor because they just use off-the-shelf control ICs, yet failure in any of those is as bad for the availability of the system than the failure in the big inverter power stage.

[ogden]

If you say that dozens of small supplies have same reliability as hi-power supply of motor controller, then you are not doing right design decisions.

[wraper]

If you say that dozens of small supplies have same reliability as hi-power supply of motor controller, then you are not doing right design decisions.

Wut? Why they should be more reliable or less reliable? Not to say it's hard to estimate what will be the real weak point in the design unless you knowingly cheapen out something too far. Regarding to safety, if your design bases safety on measuring ripple on capacitors, you're a moron. If it's say used in elevator motor driver, then safety is achieved by completely different means. Like contactors disconnecting the motor if abnormal condition is detected, breaks activated and similar. The design must be inherently safe.

[ogden]

If you say that dozens of small supplies have same reliability as hi-power supply of motor controller, then you are not doing right design decisions.

Wut? Why they should be more reliable or less reliable?

Different types of capacitors maybe? Small point of load supplies can happily live with ceramics chip capacitors.

Regarding to safety, if your design bases safety on measuring ripple on capacitors, you're a moron.

LOL Thank you for pointing out misuse of term (electrical) safety device. It is failure detector at best.

[bdunham7]

I am looking at designing an early detection sensor for the detection of failing capacitors in...

My sensor would detect the audible noise emitted during the venting process and when detected...

To me the main issue seems to be that your proposed method hardly qualifies as 'early'.  Secondarily, properly specified and designed systems don't usually have the capacitors vent as a common failure mode.

[Siwastaja]

Indeed even when the capacitors fail, actually venting forcefully enough to make sound is extremely rare, that would mostly happen with grossly undersized parts, wrong polarity assembly, etc., and noted within minutes of post-assembly testing.

All capacitors I have ever replaced have either died just bulging a bit, or some have slowly oozed out the electrolyte during months or years.

[Ian.M]

Morwind hasn't been back:

morwind
Newbie
Offline
...
Posts:1 (0.167 per day)
Date Registered:February 06, 2021, 03:40:36 pm
...
Last Active:February 06, 2021, 04:00:21 pm

not even to read the first reply in the thread (my one).
Of course its possible they are lurking, but that would imply they do *NOT* want to interact with us, which frankly wouldn't be surprising given the probable younger agegroup ("degree student") as they may well be misinterpreting the fact that their concept has been shot down in flames as hostility towards themselves.

@Morwind,
If you are lurking *PLEASE* log in.  We value full and frank technical discussions and are eager to help.

[MathWizard]

Old thread but I was just powering back up a stereo I'm working on (speakers are off), only applying ~75% of mains voltage from the auto-transformer, and I hear a hissing noise like air escaping, or like strong solder flux boiling away when the iron hits it. It lasted about 1-2seconds, and faded away. So I turned it right off.

The voltage rails looked like they were bouncing around some, and I powered it again, and hear the same noise in a few moments.

I had already checked most all the all the caps, out of circuit with an LCR meter. I put them all back with the polarity as marked on the PCB.

I've never heard a sound like this from a circuit before, so now I have to take it all off the heastsink again, and get covered in TIM again, and strip the screws/mounts even worse....