Faking a proportional valve

[parky]

I don't know the application you have but going back to the late 90's we used some propotional valves with built in pid controllers, we did not use the PID that was done in the PLC but essentially these units pulsed the air supply you could here them they did not have to be really fast and it was essentially a 5 port valve centre off (center for the US before we get hundreds of posts on spelling). These worked very well on a number of in-line temperature control. Indeed I did some mods on an existing machine that used a servo drive on a sealer, the base holding the packs was sent up towards the head, stopped a given distance before the sealing head to inject gas into the packs then carried on to full compression for sealing & cutting. The servo's were expensive, actualy were more troublesome than the original pneumatic system on the early models, I put back the cylinder (a very large one) with a LDV (I think it's called) basically a linear encoder type unit then the program was pretty simple to run at full air til gas position minus x mm then an analogue flow control driven from the PLC slowed down until position reached, the final compression stroke was full bore. this proved far more reliable in regards to the wet & harsh environment, cheaper, the only thing was the cylinder needed regular maintenance i.e. seal replacement & greasing perhaps once a year. these machines ran for over 20 years with no problems so in this instance the old tech was cheaper, far more reliable & easy for the Maintenace teams to understand.

 

[Tinine]

PWM duty cycle, % ≠ spool position, %

And lets not even get into valve characteristic, though of course that can be modeled.

If the speeds are linear, the frequency probably contributes more than the duty cycle. The only way this works is either with spool position feedback or non-linear actuator-vs.-spring characteristic (model/feed-forward).

If we are talking closed loop then I would treat it like a servo-drive in torque mode, which I use exclusively.



The drive, apart from commutating, becomes a transconductance device, proving current on command and my PID takes care of velocity/position same approach (although crudely) could work with a bang-bang valve. Yes the deadband is horrendous but we have techniques for handling that.


In my case, I am retrofitting a 3-axis CNC tube former but it really irritates me, the fact that the program loads instantly but now the set-up guy has to run around the machine, adjusting manual flow controls using his little chart. Applying a 1KHz PWM with whatever percentage duty-cycle eliminates (or will do)this time-consuming task.
Part-runs are getting shorter and changeovers more frequent so they will definitely soon see the benefit.


Rgds,


Craig

 

[MaxK]

What are the system requirements?

Not high pressure, relatively slow valve, not high demands on control accuracy ( #16 by @parky)?

OR

Giant pressure, ultra-fast valves, extremely high demands on the accuracy of valve control (the system described by @Peter Nachtwey - if I understand correctly, the valve is controlled along high-order curves to reduce n-order derivatives in the system)?

 

[Tinine]

What are the system requirements?

Not high pressure, relatively slow valve, not high demands on control accuracy ( #16 by @parky)?

OR

Giant pressure, ultra-fast valves, extremely high demands on the accuracy of valve control (the system described by @Peter Nachtwey - if I understand correctly, the valve is controlled along high-order curves to reduce n-order derivatives in the system)?

There are sooo many simple applications out there where an actuator needs to reach a target with reasonable accuracy and then immediately retract.
So there might be a PLC with a +/- 10v output and equipped for some form of feedback. This is where they tend to use spool-feedback proportional valves. They approach the target and within a certain range, slow down to stand a chance of hitting the target. By using a PWM output, via a driver, the solenoid coil current can be modulated to regulate the flow of a bang-bang valve at a fraction of the cost. Non-linear? Oh totally but that's only a matter of software fiddling



I believe I have a rough video of me demonstrating this...where the heck did i put it


Rgds,


Craig


Edit: Oops, just realised that you were addressing Parky.

 

[Tinine]

Ah-ha, on my DropBox


Proportional Bang-Bang, totally open loop






Craig

 

[parky]

Well my point was not knocking servo's although many apps that used to use simple control have now been replaced by servos for no reason other than it's high tach makes me look good.
we had loads of machines where things originally done the old way, the newer ones (I'm talking about 10 or more types that the new machines have servo's).
so in essence although accuracy is only needed to +-1mm they have supplied these machines with ther servo's, there are many enhancements but the servo's are not required as the older system was & still is robust, the original used a sensor (some 3 sensors) for gas positions (depending on pack) but I replaced these with the LDV feedback for positioning. Over a period of 4 or more years we had little trouble with either however, when one went wrong for any reason the pneumatic old system would be up & running within half an hour, the servo's in most cases this took at least a whole shift or even a few days. The speed of both systems are limited to factors of the pack and contents so even if the servo's could do it faster it would not work (we tried it).
So it's like how people crave the IPhone yep pay £1000.00 plus because it's an IPhone, my samsung was £120, it does all the jobs i need it to & I don't think the IPhone has much more to warrant the cost.

 

[Tinine]

Well my point was not knocking servo's although many apps that used to use simple control have now been replaced by servos for no reason other than it's high tach makes me look good.
we had loads of machines where things originally done the old way, the newer ones (I'm talking about 10 or more types that the new machines have servo's).
so in essence although accuracy is only needed to +-1mm they have supplied these machines with ther servo's, there are many enhancements but the servo's are not required as the older system was & still is robust, the original used a sensor (some 3 sensors) for gas positions (depending on pack) but I replaced these with the LDV feedback for positioning. Over a period of 4 or more years we had little trouble with either however, when one went wrong for any reason the pneumatic old system would be up & running within half an hour, the servo's in most cases this took at least a whole shift or even a few days. The speed of both systems are limited to factors of the pack and contents so even if the servo's could do it faster it would not work (we tried it).
So it's like how people crave the IPhone yep pay £1000.00 plus because it's an IPhone, my samsung was £120, it does all the jobs i need it to & I don't think the IPhone has much more to warrant the cost.

Guilty as charged
If it moves, I look for a reason to stick a servo on it


Thinking of getting a cobot (collaborative robot) for the house, just cuz


Craig

 

[parky]

Yeah... My mate purchased a robot lawn mower, It failed to get all the grass cut, usually ran out of juice a number of times before it did a reasonable job, did not collect the grass, got stuck on a dogs toy & took hours to do the lawn. Would have been better to buy an electric powered roller one & tie it to a steak in the centre of the lawn.

 

[drbitboy]

ah. The PWM is controlling velocity.

I missed that .

Something else (position control?) is setting the velocity setpoint.

 

[Tinine]

ah. The PWM is controlling velocity.

I missed that .

Something else (position control?) is setting the velocity setpoint.

No feedback at all in this example. The current supplied to the coil is dictated by the PWM's duty cycle on a 1KHz carrier. The spool is spring-return-to-center so the solenoid is acting against the spring. The greater the duty cycle, the greater the flow. I have another cylinder here with a linear potentiometer but the machine isn't quite ready for powering up yet. I want to see what I can do with that and also track down my string-pot (aka: Yoyo) incremental encoder to test further.


The solenoid coil is nomally driven by a SSR but for that test, I substituted a LMD18200 H-bridge ($8 aliexpress) module that I just happen to have plenty of. A single H-bridge is a neat solution because it can drive both coils without having to mess with the existing wiring.


I know of so many machines where the bang-bang valve shuts off when a striker hits a limit switch. If they want to improve accuracy, they slow the whole thing down. If they want higher production rates, they speed up but repeatability is all over the map. This approach would solve the problem.


Rgds,


Craig

 

[PreLC]

Good controls for bad mechanical system =
Bad controls for good mechanical system =

 

[Tinine]

A big problem is that you don't have spool position feedback like the servo quality proportional valves do.
In simulations it didn't work. I couldn't get the valve to switch fast enough. If it won't work under ideal situations, then it won't work with all the uncertainties in a real system. The PWM frequency needs to be high. What you are trying to do is shift mechanical complexity to electronic.

So, apart from the keyboard, you clearly are clueless about the real world. Yup, you types are everywhere. Totally useless when something actually needs to get done.


Sussed

 

[PreLC]

So, apart from the keyboard, you clearly are clueless about the real world. Yup, you types are everywhere. Totally useless when something actually needs to get done.


Sussed

Serious question, were you just looking for validation from the forum members? Cause your system obviously works(At the moment, and depending on the mechanics, perpetually), and critique is just reverted with personal attacks.

I deem you validated. Other readers beware, ymmv.

 

[Tinine]

Serious question, were you just looking for validation from the forum members? Cause your system obviously works(At the moment, and depending on the mechanics, perpetually), and critique is just reverted with personal attacks.

I deem you validated. Other readers beware, ymmv.

Ref: My OP which states that I'd already experienced success with this approach. The video was from a couple of years ago when I first had the idea. It wasn't implemented on that customer-machine because I wasn't there for that reason. I had a H-Bridge driver with me and so I took advantage of the fact that I had a real machine to play with to test the theory. I literally spent minutes with the code and it worked from full flow down to ~1mm/second.


Today, I am on a project that can actually make use of this but I'm not even at the power-up stage, yet. This is what prompted me to post.
In a previous life, I shipped ~200 Vickers Proportional valves/year and this solution would've achieved the same thing, saving me $400K+/year.
It would be cool if someone here could realise something similar.


And then up-pops a theoretician to inform the thread that what I have already proven, is actually impossible. Now I regard that as an "attack".


There's never a shortage of "can't do" types.


Craig

 

[Peter Nachtwey]

And then up-pops a theoretician to inform the thread that what I have already proven, is actually impossible. Now I regard that as an "attack".

Thin skinned you are.
It was not meant as an attack and I said it won't work, at least not for the applications we do. Moving unloaded cylinders in open loop is one thing. Moving tons with speed and precision is another. The reason why it won't work is that you can't switch the spool back and forth fast enough to keep severe pressure spikes from occurring. If you expect the spool to float at some point in between then the spool has no way to keep from being moved due to flow forces.

So why are your panties in a bunch? You have your market, and we have ours. Ours is worldwide. We compete with the big boys. We swim with the sharks.
We have done few pipe bending machines probably due to cost and we have features that most pipe benders don't need.

You are the one doing the attacking reducing me to just a theoretician.
I am much more and have done much more.