Determining Solenoid Condition

That's all very interesting, but any solenoid I've seen go bad it was an open circuit in the coil, or some trash in the line that got caught in the plunger keeping the valve partially open....I don't think I've ever seen a spring wear out....Wouldn't a better approach be to just check coil temperature?...or coil resistance?

In any case I found these links that may be of interest...

https://www.nxp.com/docs/en/data-sheet/MC33811.pdf

http://www.ti.com/lit/ds/symlink/drv103.pdf

https://www.nrc.gov/docs/ML0403/ML040360191.pdf

These are not solenoid valves,they are a cylindrical solenoid operated "piston" shaped armature that is spring return.

There has never been a coil failure over millions of operating hours,they are inherently current limited and overload protected.

In normal operation they are activated for only about 100 milliseconds,with a duty cycle of about 10%.

The problems occur when the solenoid becomes jammed and product does not dump,creating a "log jam" behind it.

There is currently no feasible method to determine imminent failure of an armature "piston" that actuates or returns too slowly.The purpose of this proposal is to know when an armature is slow to return or engage,and schedule a replacement during a maintenance window.I am sure that a hall effect or optical sensor would work,but the expense would be too high.

A shielded pair or coax fore each gate signal would be better.

It might even be practical to monitor the signal at the controller source ,but each solenoid would have to be individually calibrated to compensate for cable length,resistance,etc.

Theoretically,a new solenoid could be characterized when installed, by energizing it,and these measurements would be used for future reference.

I am sure there are experts that could create the software required,but my main concern is the hardware at this point.

Theoretically,a new solenoid could be characterized when installed, by energizing it,and these measurements would be used for future reference.

I am sure there are experts that could create the software required,but my main concern is the hardware.

Thanks for your feedback on this,and I will review your links later as I have prior commitments to attend.

Sounds to me like you are focusing on the wrong part. With over a million hours of operation and proper circuit protection these coils continue to be sound. Monitor either the dispensed product (weight, image, optical switch) or possibly this return spring timing with the addition of a fast switch.

The coil has never been the problem,and the mechanical portion is the part I am interested in.There are many ways to ensure the proper speed and return of the solenoid,but I am looking for a simple relatively inexpensive method.

Adding switches and sensors will require lots of labor costs as well as material costs,so I am hoping to find a way to analyze the existing wave forms for information.

I hope it can be done at the controller that generates the firing signal to the solenoids.

If the demo proves effective, a modification to the driver board could be implemented to incorporate this feature.

If a solenoid fails either due to a weak spring or otherwise, and whether an alarm sounds or not, it appears that the logjam will still occur and need to be cleared.

Perhaps a better approach may be to monitor product movement at each point and alarm and/or stop the conveyor if that movement slows or stops.

This monitor may be able to detect slowed movement of product due to slowing solenoid release even before a total failure occurs and could be set up to alarm only on a slow down and stop the conveyor upon a total jam.

Often a slowed release causes a following item to enter the gate and get stuck by the closing gate thus causing the jam, an approach to overcoming this may be to increase product separation on the conveyor.

I'd be surprised if you can see any difference at all in the voltage flyback waveform. You're confusing where and what kind of energy is stored in each component.

(From the above link)

This is the classic description of how a flyback voltage is formed and is very good at teaching the concept of flyback spikes. However, it suffers from being unrealistic. If one were able to achieve this circuit the energy stored in that 10 Henry inductor will have a time constant of one quarter of a picosecond. The parasitic wiring impedance in parallel with the inductor will surely be smaller than 40 tera-ohms. What is accurately portrayed in the linked article but not made explicit is a relaxing (discharging) inductor is a current source and not a voltage source. The voltage waveform will be determined by the coil and the discharge path impedance. Keeping the discharge path impedance fixed for waveform comparison is critical.

The crux of the problem is time. If one slows down the inductor discharge path time constant low enough to see an inductance change as the slug is moved by the spring the peak discharge voltage will be so small it will be very difficult to measure.

Also by slowing down the coil release the magnetic force on the slug will also be gradually released. The non-linear factor of stiction may not be overcome by the spring. Thus by trying to measure for a sticky spring you may make it more likely to stick.

I'm understanding that you are primarily interested in how fast the plunger retracts when current is removed from the solenoid.

The voltage across an inductor is equal to the rate of change of L times I, V=d(LI)/dt, where L is the inductance and I is the current. Normally L is constant, but in this case, when the plunger moves out, L decreases as well as I decreasing. In addition to this, if the plunger is magnetized (which it would be), there would be a voltage generated as it moves out of the solenoid.

So, bottom line, I would expect that you would see a difference in the waveform depending on the speed of plunger retraction, but you need to put a scope on it to see if it's significant or not. I'm thinking you will see a larger inductive kick if the plunger retracts more quickly.

If you decide to use analog multiplexer IC's (4051B or relatives), be sure that the circuit is biased so that the voltage waveform stays between V_DD and V_EE of the analog multiplexers.

http://www.ti.com/lit/ds/symlink/cd4051b.pdf

I also would expect a difference in the waveform when the speed of retraction or actuation changes,and I intend to look for this at my next opportunity.

Matter of fact, I am counting on this change in waveform to help determine a weak spring or slow actuator.

Thanks for your input on this.

You did not write if the solenoids are DC or AC. I suppose DC, because you mention a current drive. What voltage? - affects voltage rating of multiplexers.

Obviously, 4051s could be usefull - but if all solenoids are not on together, you may need only a small number of shunt resistors on supply feed side, it might be possible to detect faults like one solenoid driver always on.

A current drive will not have infinite voltage, so there will be an exponential rise.

My main worry would be that mechanical delays/times of movement may be far longer than any R-L delays, so you will not see any effect on current curve. DC solenoids I have dealt with can have very long release times (seconds) due to diode & shorted-turn effect of aluminium casings.

You need to do some tests to determine current & mechanical movement time constants before worrying about in-production measurement circuitry.

If the coil is not saturated, or is de-energised while returning by spring, applying an AC sine voltage & measuring AC current may reveal inductance, in effect the magnetic gap/plunger travel. This appears capable of detecting incomplete or slow return.

You have not clarified if stuck-open or stuck shut is most costly.

The solenoids are driven by a 48volt dc pulse,and the return portion is more liable to be slow than the active.

A delay in either direction can cause a jam,but delayed close (spring return) is worse.

These are very fast solenoids with a very short travel distance.

The total cycle time has to be less that 100 milliseconds.

The application of a very low level AC signal to detect the health of the solenoid is a very good idea,and I will explore this possibility.

It could possibly be capacitively coupled to the solenoid to provide real time information.

Thanks!

Detecting any difference in supply feedback due to slow retraction is IMHO a non-starter. If a potential problem, initial design should have included end of travel sensors and a timer.

I agree it would have been nice to have those features,but I am forced to play with the cards I have been dealt.

I cannot just throw up my hands and walk away because they are not designed as I would have done it.

When these devices were purchased,I am sure it was a competitive bid process,and the cost was a great factor in the winning bid.

You cannot argue technical details with an accountant.

The magic words to argue with an accountant over technical details are "maintenance costs" and "critical downtime". How one uses these magic words is different with each scenario. They don't always work. Nothing always works. But if one does not attempt to cast this spell...

Unfortunately accountants are taught to focus on quarterly returns,not long term costs.The theory is if you take care of the quarterly returns the annual return will take care of itself.One of the reasons the Titanic sank so quickly was because the original design was modified.There was no need to take the watertight seals all the way to the ceiling,because the water would never get that deep anyway.And no need for so many life boats,because the boat would never sink.

The engineer yielded to the pressure of the accountants,and went down with the ship out of guilt.

I think the back emf can be sensed / picked up by a hall effect sensor since the di/dt is infinitely high in that application... L di/dt of the coil.. I believe hall effect sensors response time may possibly match the switching speed of the coil?

I'm guessing the di part of the emf in each coil to be very high while its decay time is very short almost zero?

Thinking of feeding the sensed signal to an Arduino for ADC processing may be a possibility also..

Good luck.

Maybe this:

A small transformer at the controller end in series with the solenoid. Characterize the signal in normal operation, then hold the plunger out and see how the signal is different. It may be as simple as detecting a level change. Feed the good/bad conditions to some muxer to alert an operator, etc. If you want to get info re a weak spring, you might need some software to detect waveform change.

Don't hypothesize on hypotheticals.

Hook your CRO up, get good waveforms, and bad waveforms. You may have to simulate a sticking solenoid to get the bad?

The answer, then, may be staring you in the face.

My intention from the start was to do just that,but the problem is how to detect and differentiate the good/bad in an economical way.

The simplest device I could think of is a very low value resistor in series with the reverse-biased diode,and use the voltage drop across the resistor to generate the signal.

Each group of 32 solenoids has a driver board and tracks each individual product by photocells as well as driving the solenoids.

If a photocell is blocked more that 100 milliseconds,it is considered a jam,and the machine stops.

The inertia of the machine causes a delay in the actual stopping of motion,and this creates a further jam behind the initial jam.

The previous board then loses tracking data,and the jam as well as several products behind the jam have to pulled off and be run again.

The piece that created the jam is usually damaged,as well as some of the following pieces.

The total cycle time of the solenoid,in both directions has to be less than 100 milliseconds.

A good solenoid does this with time to spare,so detecting a sluggish solenoid before it exceeds the parameters would save time,reduce damaged product and increase run time.

Not all jams are caused by the solenoids;sometimes it is a product that is dimensionally out of spec: ie:too thick,too long,etc.,tailgating by the following product(inadequate spacing).

For instance,consider potato chips,where the bag is filled with nitrogen, to prevent crushing of the product.

An overinflated bag will jam,but this is not because of the solenoid,it is a result of another process further upstream.

Likewise,an underinflated bag will cause a jam because of improper sealing,etc.

These problems are more easily resolved,but the sluggish solenoid is more subtle.

Of course,when a solenoid generates more jams than it's neighbors,it is suspect,but this is a reactive response.

I want to be predictive and proactive.

Right now,I am trying to eat this elephant one bite at the time.

Eventually I will get there.

Thanks for your feedback on this!

Remember the time constant equation for an inductor is τ=L/R. A very low value R will lengthen the amount of time it takes for the solenoid to turn OFF.

I have ruled out a resistor,and I am pursuing the use of a low level AC voltage to monitor the armature position,as suggested by 67 Model,post #18.

If I capacitively couple a low current AC signal to the solenoid,it will not be seen by the DC driver voltage,and it is possible I could use the result as a real-time monitoring circuit.All solenoids could be monitored and trended for condition.

I realize the current should be low enough to prevent "plugging" of the solenoid and creating a problem.

Thanks for your feedback.

This is the "back of ciggy packet" picture for measuring solenoid impedance. AC Signal fed to sol via 4051 analog gate, with 10k/1 mmF coupling for each sol. The flywheel diode will not conduct enough to affect measurement. If you really have current drive to sol, that means drive transistor is Hi Z, if not "on" resistance of saturated transistor of 1 ohm makes measurement while ON impractical.

Imponderable is eddy current effect in solenoid - DC solenoids have solid cores. Need for testing. 60Hz is bad news for interference, just easy first try. Say 100 Hz, with synchronous detector, keeps away from 60Hz & hamonics. Higher frequency will give more signal, but more eddy current. A brutal approach with square wave might show effect of frequency quickly.

Here, for trial, I would use Maya44 USB 1v rms to 18 bit audio interface with Sigview software to give "scope"/FFT spectrum & pick out wanted signal from any interference. You can use the audio input of a PC at a pinch. Obviously, protect Maya input against 48 V switch spikes.

67model

How come you draw circuits upside down?

Maybe he is left handed?

If so,at least he is in his right mind.

I like your idea about using AC to test the solenoids

The solenoids could be checked during the "off" interval.

Plenty of time since they have a very low duty cycle.

They could be checked on every cycle.

If the voltage is below the bias voltage of the flywheel diode,it would be "invisible" and not be a problem.

The best frequency is yet to be determined by trial and error.

Thanks again for the ideas.!

I can recommend SigView for "proof of concept". It can trigger a scan of a specified no. of samples from a second, trigger, channel, even delay after trigger or record before trigger in buffer. I guess the trigger could be the solenoid drive on or off - so it could be same channel as waveform samples. It can then calculate the rms or other value of the wave during the scan and push the values into a .csv text file. A spreadsheet can determine the scatter or standard deviation of the values in the file or successive parts of it [or Sigview itself can swallow a .csv file and produce a probability distribution or histogram of the values in that file]. Both good & faulty valves could be tried on the "bench".

I have several times used a laptop tied to a supply (including mains neutral) by its chassis or microphone input , with laptop on a wooden table. The laptop PSUs have small coupling capacitance from line & high isolation voltage - 4 kV rms I believe, so it does not upset the measured kit.

Further to posts #27 & #31, below is circuit for connection to laptop microphone socket....

Scope wave form with L as solenoid(24V secondary of "plugtop" transfo, about 0.05 Henry) - the "sine" source is actually a sawtooth from the capacitor of a "555" type R-C oscillator amplified by an LM386 to 4V peak to peak. The solenoid switching is simulated by a 5 Hz square wave, of course the 1mmF/10k, with 10ms time constant only passes a spike......

This is the Sigview recording in triggered mode, the exponential decay of the spike affecting the 100Hz measuring wave can be seen - the spike is clamped at full scale of mike input A to D converter......

This is a zoom into the wave after the exponential decay.....

Finally, this is a fourier spectrum of the zoomed part.....

67model

Way out of my league but I give you a GA for effort.

If spring weakness is the cause of failure (the only cause?) would the force needed to compress the spring be a reliable indicator to detect early failure - and something easier to measure?

Alternatively, to my mind, energising provides the prime control function - with spring return being secondary. If spring return is also critical in the process, then it can be brought more under control by using a solenoid return.

My earlier experience, in pneumatics, 'stickiness' in the valve caused problems when relying on the spring for action. No failures occurred when replaced by air operated armatures.

Our own naive attempts to increase the return force, by adding a few packing washers to overcome the stickiness failed because the solenoid was then not strong enough to compress the spring.

Solenoids and springs are balanced and matched pairs - so we found out.

I mention it for what it is worth.

Spring weakness is not the only cause of a slow return.

Mechanical wear and contamination by dust,etc. can also cause the problem.

Pneumatics are too slow for this application,even when using a quick exhaust valve.

This application doe not require proportional control.

It is simply ON-OFF.

In pneumatic operated valves,a valve positioner was used to ensure that the valve moved to the desired position,however,these are relatively slow compared to these solenoids.

Thanks for your input.

I've been thinking over the flyback signal you could expect to see and am having some problems. I'm making a few assumptions which may or not be good. First, I'm assuming that the solenoid plunger and any pole piece is ferrite so that it loses magnetism quickly. The flyback curve will be for a ferrite core coil until the plunger moves. Once the magnetism level decays to the point that the pole piece and plunger separate, the impedance of the coil/core system will change based on the residual remaining magnetic field and the plunger displacement within the coil.

Based on that, the difference between the strong spring and weak spring condition will only be from the point where the plunger separates and displaces from any pole piece and the magnetic decay characteristics of the plunger and pole piece. This would be at the very tail end of the de-energization of the coil and would be driven by residual magnetism in the ferrite and the change in impedance of the coil. I would estimate that it's going to be pretty tiny by comparison with the initial de-energization signature. I don't envy you the task as you will also need to discriminate among potentially several parallel wired solenoids on the same circuit.

I'm wondering if you could float a very low amplitude, high frequency signal on the circuit, measure the time and impedance of the coil directly and look for the impedance change when the pole piece and plunger separate. This doesn't help you with which IC chipset you need for processing the signal, but I would guess that whatever you end up with is going to need a fairly high clock speed and a fair amount of intelligence.

Personally, I would bail out on this one and get solenoids with remote position indicator switches and monitor the position indicator signal on a time base vs DC power signal and look for the time difference there.