posted August 21, 2002 12:32 AM               

I was wondering if anyone has ever experimented with, and/or heard of another method of control?

Two reasons for my question -

1). solenoid valves (especially on our APET extruders) take a beating, and require frequent maintenance. A control scheme with better reliability would be a good thing ...

2). it seems to me that more even cooling control could be obtained if water flow were varied from 0 to 100% rather than pulsed.

posted August 21, 2002 07:03 AM            

The question you should be asking yourself is "why do I need so much cooling?".

If your extruder is running isothermally you should need minimal cooling. The reason for your need of cooling has to be that your heater zones are overshooting due to adiabatic running of your extruder.

Where necessary we use air cooling (fans) controlled using PID controllers.

Regards,

posted August 23, 2002 06:36 AM               

With water cooling, the intial pulse of water flashes to steam and removes large amounts of heat quickly, as the band cools, less heat is removed.

With On/off heating and cooling, the cooling cycle is going to be a lot more effective than the heating cycle. Some controllers have a gain control to compenstate for this.

Are you using a closed loop system? You should supply water to the barrel cooling elements at the pressure recommended by the maker of your extruder (normally 30PSI)

Steve H

posted August 24, 2002 05:16 AM               

Let me preface this by saying that, IMHO, extruder water cooling is one of those areas that is deceptively simple, and does not appear to be very well understood once one delves elbow-deep into the complexities.

I've seen mostly smaller extruders using forced air cooling, and only a few larger ones (and those typically because the polymer did't take kindly to rapid changes in temperature - the ones I'm thinking about processed nylon). In our plant forced air would be a hard sell, if only because the exhausted air would tend to heat our alreay toasty warm plant (and the additional cost of ducting it through the roof would not be looked upon favorably). That isn't to say it not a bad idea ;

Our machines indeed have closed loop systems using distilled and treated water, and, at least in the system I'm currently looking at, feed side manifold pressure runs in the 32-35 PSI range at 110 degrees F.

This particular extruder has a vendor-supplied SCADA system that brings back temperature, drive RPM/% load, and various pressures, and can download these values to a DIF formatted file. Unfortunately, it does NOT log temperature zone % output (which would be very helpful in this case).

Barrel zone 1 trend showed almost sinusiodal oscillation with a range of about 18 degrees F (setpoint of 615F, +/- 9 degrees). Melt temperature was ranging between 548 and 552 degrees F.

The temperature controller is a Eurotherm 94C with high and low cutback terms set to 'auto', P=35, I=750 seconds, D=140 seconds. I tried several controller-based manipulations, including widening the prop band, and increasing integrating and derivative time with little change noted after a hour of 'settling' time. Looked a little further, and found the cooling gain had been set to 1.0. Normally, I set this to at least 0.5, and sometimes lower, but altering this term to 0.5 improved control only slightly (to about 14 degrees peak-to-peak).

Sliced the next layer of the onion, and checked the flow control needle valves. Found these allow 6-1/2 turns from full close to full open, and all but one zone was cranked fully open.

When I closed zone 1 flow valve from fully opened (6-1/2 turns) to 5/8 turns open there was a "night-and-day" difference. The temperature trend showed oscillation stopped within 15 minutes of this change, and stayed essentially flatline (about 2 degrees F peak-to-peak, and fairly random, non-forced oscillation). When the wild swing were over controller output ended up in the -5 to -15% range.

Opening the valve for even slightly more flow (3/4 turn) and sinusoidal oscillation returned (although at a lower peak-to-peak value), and closing it slightly more (1/2 turn) allowed a monotonically rising temperature for the 1/2 hour I allowed this 'experiment' to run.

This indicates the valve position that allowed proper zone operation allows only slighly more flow than the position that would not have allowed enough cooling, and also that all or nearly all of the water volume was flashed to steam.

This is where digging deeper into the metaphorical onion becomes more difficult ... I'm envisioning that the typical cast-in heater cooling circuit tubing is single serpentine (rather than a network of parallel tubes). The tubing OD is 0.500", and, after playing around with some numbers cut from whole cloth - figuring a 24" long heater with a 6" ID has the tubing bent on 3" centers, placed it in the center of the casting (which is 1-1/2" thick), and other highly laughable assumptions - this heater holds on the order of 1/2 gallon of water.

Years ago I profiled the characteristics of a similar needle valve on a different extrusion system (except this one was 9 turns open to closed), and learned flow is fairly linear with valve position until about 2-1/2 to 3 turns of approaching the closed position, at which point flow becomes non-linear. This study was done many years ago (circa 1988), and I haven't tracked it down yet (and may never, and end up re-doing it).

Flow will be dependent on the pressure differential between feed and return manifold, the flow control valve setpoint, and controller % output (controlling the on-off solenoid valve). How much thermal energy is pulled out of the heater will also depend on what fraction of the water submitted to the heater flashes to steam, because (as Steve points out) the phase change drops more energy than if the water simply reached, but did not exceed, 212F, and, to some extent, the entry temperature of the cooling water.

Musings

* Lower controller % output (especially when coupled to lower flow valve settings) will tend to allow a greater fraction of the water to turn to steam.

* The PID loop will tend to be more stable if cooling is mostly in one phase (liquid or steam) or the other, but becomes more problematic when mixed phase due to inherent non-linearities at the phase transition. PID doesn't handle non-linear all that well.

* Since the one guaranteed phase is that at least some of the water will turn to steam, then I'm wondering whether it would not be better to meter coolant in such a way as to guarantee it will all turn to steam. I'm thinking of the example of a refrigerant system expansion valve. There are pitfalls to be overcome, but it may turn out they aren't as intractable as they at first appear.

* Typical extrusion control systems only turn on and off flow to the heater cooling jacket, and leave the setting of the flow valve to chance. In my experience, machine operators and process techs tend to open them up 100%, and no amount of explanation seems to break this pattern. I'm figuring servo/stepper technology has come along far enough that reasonably priced servo-controlled needle valves are possible.

If so, it seems to me that closing this end of the loop would be beneficial.

Hence, my wondering if anybody has ever attempted anything other than the current defacto method of controlling cooling in water jacketed cast-in barrel heaters.

posted August 24, 2002 09:02 AM               

I have also found that adjusting the time base for cooling (the period of time between consecutive openings of the solinoid) to be very helpful. If you can set the time base to something like 20 seconds, the maximum on time to 2 seconds, proportionatly control withing the 2 seconds, and set the needle valve to a faily high flow, you can keep the control in the "flash" zone. This will give you much more control.

posted August 24, 2002 09:39 PM               

Great post, seriously I'm in awe. As I understand it, the non-linear cooling effect of evaporative water cooling can cause severe temperature oscillations. I belive that some suppliers of temperature controllers have developed an algorthim that is specfically designed to cancel out the inherant non-linearity of this type of cooling.

A paper was done by John Radovich Jr of Davis Standard- I don't have a date but he explores the heat transfer mechanism involved in barrel cooling (both water and air).

Despite the complications of plumbing, water is an effective way to remove heat from a barrel. Particularly in an air conditioned building where air cooling would adversely affect the heat balance. Water cooling is also recoomended for processing polymers that bridge easily or are very shear sensitive.

The contracool system developed by Battenfeld Gloucester, whwre air is sucked in and down over the barrel elements and then ducted away (outside) would be an option for a plant that has AC.

posted August 30, 2002 10:42 AM               

When time allows I'll give D-S a call, and see if John will forward a copy of this research. I'll keep the contracool system in mind for new installations.

Tom - I'm not quite tracking your thought, so bear with me while I restate it somewhat.

A controller using a 20 second timebase, will be on 1 second at 5%, 2 seconds at 10%, 5 seconds at 25%, 10 seconds at 50% output, and so on.

Some controllers have output limiting terms where, for instance, a limit setpoint of 20% will allow output to change within the range of 0 to 20% regardless of whether PID calculates higher output values. I'd have to check this particular controller, but don't recall if the heating and cooling outputs have their own output limiting factors, or if a single limiting factor is shared.

If the former is true, then setting up cooling for output limiting at 10% with a 20 second timebase would provide the action you describe. I'd never thought about using output limiting in this way (thanks for the tip), but it would certainly keep the heater within the flash phase.

Its interesting to find another technique for forcing cooling to stay in the flashing regime that appears to improve temperature stability. If this is a general observation and holds true in the majority of cases then my original thesis (that maybe it's time to re-think the barrel cooling control paradigm from the ground-up) gains some currency.

Steve, many Eurotherm controllers (including this one) offer a selection of cooling modes. If memory serves, they are 'linear', 'water, 'air', and perhaps one other mode. The instruments on this machine are set up for 'water' cooling.

It does bring up a rather sore point (with me, at least), and that's the paucity of hard, detailed information on what differentiates them, and other 'non-standard' terms. I suppose to a great extent this is because a manufacturer has legitimate intellectual property concerns, and a very detailed explanation would let enough of the "cat out of the bag" so others could simply replicate their functions. Still, there are aspects of controller set-up that, because the underlying principles of how particular terms are used are relatively undocumented, bring me in mind of being a sort of algorithm-casting witch doctor.

Tom, Your point about the location of melt film formation, and the much greater thermal flow through melt film as opposed to that of unmelted pellet and reclaim feedstock is well taken. There are a number of variables that affect this, including but not limited to the following:

• Screw speed
  
• Barrel temperature profile (including feed throat)
  
• Screw temperature
  
• Ratio of pelletized to reclaim material.
  
• Grading (size distribution) of reclaim material.
  
• Pellet geometry (strand cut vs. round)
  
• Pellet size distribution
  
• Feedstock temperature and moisture content
  

I am certain variations in at least some of these are related to the overall instability of the extruder when running this particular product. For one thing, this extruder uses a melt pump, and thus changes screw speed to maintain suction pressure. There was significant instability in suction pressure and thus also screw speed.

This product requires fairly high #/hr throughput (probably the highest we run on this line), and (if memory serves, and this extruder has an 11.5:1 gearbox - 1150 RPM base speed/1750 RPM fully weakened ... 152 screw RPM maximum) the average speed was about 82% of full speed, and maximum observed speed was 94% of full speed. Armature amperage averaged 95% of FLA, and maxed out (mostly during acceleration phases) to 115% FLA.

In addition, we currently run un-dried APET material, and pull moisture out solely by vent vacuum. The resulting i.v. isn't spectacular, but better than I thought it would be (the amazing thing about a dancing bear isn't how well it dances, but that it dances at all). I'm not privy to the particulars, but am under the impression the screw was cut specifically for this purpose.

Checked melt pressure controller PID, and rate limiting parameters, and found they were set to unremarkable values. Typically I turn off the derivative term, set integrating time for no no less than 8 seconds, and prop band in the 100 to 200 range. Significantly better operation was not obtained through their alteration. Reclaim quality, and other direct observables were unremarkable.

This nudges me towards believing things happening at the screw/barrel/polymer interface are involved.

Just for kicks, and to verify I have figured out how to embed images into these threads, here are (with any luck at all) approximately 3 hour trend charts for the following parameters.

Barrel Temperature

Pressures

Speeds

Melt Temperature

BTW - These trends were taken previous to troubleshooting, and making PID and flow changes. Also, be forewarned - some of the pressure readings are smelly (it looks like they aren't scaled correctly). Didn't have the opportunity to do anything about it ... wasn't about to kill production just so the telemetry added up.

I have no science to back me up, but the general feel of the machine is what I call the "standing tippy-toe on the top rung of a tall ladder" schema. By this I mean most everything is pushed close to the limits, so minor variations can cause disproportionately major changes in system operation.

And now, back to a restatement of the original question, "Is there a better way to control cast-in barrel heater cooling loops than with time-proportioned off-on solenoid valves?" My feeling is that there may well be, but these alternatives have not been explored.

[This message has been edited by rawelk (edited 08-30-2002).]

posted August 30, 2002 01:46 PM               

My reasons:

1) No barrel heating or cooling equipment has the ability to move heat around as fast as your process is changing. Given the instability of the process, I think the controllers are doing well.

2) Your solinoids are wearing out because you fire them too many times, because the barrel temperature controllers are tuned wrong, and/or the cooling cycle peiod is too short, and/or the process changes too rapidly.

3) The control of the extruder screw speed for providing suction pressure is way too quick. It is also likely a PID controller. I can't see that the gear pump needs 4500psi inlet pressure. Regardless the gear pump will probably function anywhere from 500 to 5000 PSI. Over 5000 PSI will proably risk breaking things. How about allowing the extruder to creep into 4000 PSI. Set the gain (inverse P) low = High proportional band. This will surely help with temperature control. Also it will reduce acceleration and over amping. You proabaly don't need any integration or derivative control. I think they might cause more problems than good here. I would also have a cutoff limit for screw speed based on experience which will keep you out of trouble. Some of these suggestions will also help the process cope with the suspect pressure spikes.

Not an easy process. Good luck.

(Great graphics, I need to learn how to do this.)

posted August 30, 2002 03:25 PM               

I presume you're extruding Apet sheet and using Welex lines. I'm not suprised the I.V. is way down. Are you preheating the feed at all? and what % regrind are you adding, is the reclaim just edge trim.

posted August 31, 2002 09:31 AM               

quote:

---

Originally posted by Steve H:
Bob
Please write a tutorial on how to imbed graphics into a post (a picture truely is worth a thousand words).

I presume you're extruding Apet sheet and using Welex lines. I'm not suprised the I.V. is way down. Are you preheating the feed at all? and what % regrind are you adding, is the reclaim just edge trim.

Steve H

---

Actually, there is an excellent tutorial on your site already.

http://www.feedscrewdesigns.com/ubb/ubbcode.html

About the only thing it doesn't mention is how to make the desired image files available.

The ISPs most people use for their email accounts grant a certain amount of space for personal web sites. Usually it's simply a matter of consulting with your ISP (because the particulars vary widely) and gaining FTP access to this area.

I use WS_FTP, but CuteFTP is also good, and there must be a couple of dozen other FTP clients that could be used. Basically, these programs allow one to create directory structures, and upload and download files to them - in effect, to use the ISPs disk space as a sort of remote hard drive.

If you wish I would be agreeable to writing a page or so to add to the existing UBB tutorial.

Your guess as to the extruder pedigree is on target. We do not now have the capability to pre-heat the feedstock, and (because this product is a drink cup) do run considerable reclaim percentages.

It's a moot point, however, because we are in the process of installing a crystallizer and dryer. When this is done I believe these stability problems will be history.

posted August 31, 2002 04:05 PM               

I've always had reservations about relying on a barrel vent to extrude undried Apet. The material is made by a polycondensation reaction where water is removed allowing the monomers to combined- heating the result in the presence of water reverses the process.
Now I'm not a chemical engineer,(just a dumb ass sparkie who's spent too much time near extruders), but I think there's a flaw in the logic of using vent extraction to process Apet.

I don't much like drying and crystallizing using heat either, but it's better (IMHOAL).

In your case, I would look at Erema's process, it can produce sheet using a 100% reclaim, the FDA has issued a letter of no objection for the use of the resultant sheet in food contact applications, the process uses mechanical energy to to dry and cystallize flake/virgin under a vacuum and the resulting sheet can have the same or a slightly higher I.V after processing.

The energy cost of producing sheet this way is 50% less than the dryer/ cyrstallizer route, it would be more than the vent extraction method- but you'd have a more stable process and a better end result.

Erema have done some work blending in chain extenders in with the feed during the cutting/drying phase and have reportedly boosted the I.V of the material.

For drying Apet Eastman recommend:

150C (300F)

4-6 Hours residence time at temp.

Airflow of 1 cfm/lb/hr of extruder output.

Dewpoint -40

So this tends to mean desicant dryers, airlocks yada yada- where Eremas idea is elegant simplicity. Now I also race yachts, and elegant simplicity wins out over awkward complexity every time at sea- which is why I like their way of doing it. And no, I don't have shares in Erema, or any connection with them.

Steve H

posted September 01, 2002 10:46 PM               

For me, it's not a question of whether solenoid-controlled cooling will work or not, because it certainly will, but rather if there are other methods not in common use that may work better.

Your point about limitations in moving heat around is absolutely valid. It seems to me there is no chance for a conventional extruder to control actual instantaneous melt temperature at a particular barrel zone.

Allow me to tick off some components in the thermal transport system, at least, so I can get them straight in my head. First, polymer temperature itself can't be truly isothermal - it will be hotter at the shear face. Of course, this is being continually folded and mixed into cooler material in the solids bed, and later, melt pool.

The thermocouple is mounted in a well drilled into the barrel wall, and sensing barrel temperature at perhaps a 3/4" depth. Melt thermal energy conducts into both barrel wall and screw, but more of it conducts into the barrel. There is a time lag for that heat to migrate to the thermocouple.

It then both radiates directly from the outer barrel wall, and (where in contact with a heater) into the element. Some of this is re-radiated from the heater surface into surrounding air, and (when cooling is turned on) also into coolant media.

There are slightly different time lags (but of the same order) between thermal energy from the heating element, and energy removal from the cooling tubes into and out of the barrel, which is then conducted to and sensed by the thermocouple (which itself manifests a time lag).

I've seen extruders with two TCs per zone - one in the barrel, and another in the jacket - and their signals paralleled to create an averaged value. IMHO, there are only a few pros, and many cons to this approach. The main pro is to, in effect, decrease measurement time lag. The cons are both process and maintenance related - temperature measurement, which is already somewhat of a lie, becomes the big lie, and this arrangement also makes for more difficult troubleshooting.

Harrel has a patented two thermocouple system that does this with more finesse, but I haven't seen one in action, and don't know the particulars. However, I suspect they bring the TCs into the control system separately, use the barrel TC as a 'master' (coupled primarily to the prop band), and heater TC to extract temperature rate of change information (primarily tied to integrating and derivative functions).

The degree of thermal resistance between heater and barrel also plays a role in both lag time, and overall thermal coupling (does the barrel heater ID mate with barrel OD well? is a thermal transfer paste used? are element clamping straps tight?).

In fact, one of our standard diagnostic and PM procedures is to remove, clean (both heater and barrel), and reseat heater elements using fresh transfer compound when gradually increasing and otherwise unexplainable instability occurs, especially in the first stage zones.

In any event, these lags mean that, by the time a sudden increase in local melt temperature is sensed at the barrel zone thermocouple, that mass of goo is already 'way downstream.

The only thing to do is attempt to keep the average temperature of the zone consistent.

quote:

---

Your solinoids are wearing out because you fire them too many times, because the barrel temperature controllers are tuned wrong, and/or the cooling cycle peiod is too short, and/or the process changes too rapidly.

---

Sorry, Tom - can't buy this (or, at least, all of it) in this instance. Very similar Asco valves are used for roll cooling control on most of our extrusion lines, and they usually hold together a year or more without requiring rebuild even at twice the cycling rate (that is, using a cycle time of 10 seconds rather than 20 seconds used on this extruder).

Also, we experience Asco valve longevity problems on extruders running APET, and, to a lesser extend, polyethylene. Those on extruders running polystyrene have less maintenance issues. The common thread for this increased maintenance demand appears to be temperature.

We've had Asco reps come in, and imagine they selected FKM fluorocarbon parts rated to 350F. Now that I'm thinking about it ... the last kit I watched get installed looked suspiciously non-fluorocarbon-based (mental note - had better check).

Standard Asco valve diaphragms are NBR (Buna N) which is a good, all-purpose material, but craps out above 180F. CR (neoprene) is usually used in refrigeration and oxygen service, and holds together to 300F. EDPM (ethylene propylene) is another good choice that tops out at 300, although it degrades if there are hydrocarbons present.

Another thing I need to check up on - not sure how these valves are plumbed. Many of our lines have a check valve between the Asco and heater to prevent steam/superheated water backflow through the valve internals, and I don't recall seeing any checks on the APET machines.

I'm going into more detail on the Curmudgeon site because these comments have mushroomed to about 3 pages (and include tables to boot), doing it this way makes it easier to author from the git-go, and then later recycle as in-house training material. The main gist of it concerns valve cycle time as it relates to MTBF.

quote:

---

I can't see that the gear pump needs 4500psi inlet pressure ...

---

This is one of the "smelly" pressure channels I was talking about. This extruder is also equipped with a mechanical dial gauge for head pressure, and (the day before) found it was broken. However, during a short downtime to deal with downstream problems it was changed, and head pressure turned out to actually be in the more reasonable 2500-2800 PSI ballpark.

There wasn't time to pinpoint the underlying problem, and perform a calibration, but my best guess (because the measured value is about twice what one would expect) is the strain gauge isolator is either set for twice the desired excitation voltage, or the max. scale is set up for a 5000 PSI probe but a 10K is actually installed, but ... who knows?

quote:

---

The control of the extruder screw speed for providing suction pressure is way too quick. It is also likely a PID controller ...

Set the gain (inverse P) low = High proportional band. This will surely help with temperature control. Also it will reduce acceleration and over amping. You proabaly don't need any integration or derivative control. I think they might cause more problems than good here. I would also have a cutoff limit for screw speed based on experience which will keep you out of trouble. Some of these suggestions will also help the process cope with the suspect pressure spikes.

---

Very good points, and thanks for the advice. I never use derivative for a melt controller (the signal is too 'noisy'), and typically use only moderate integration (12 seconds - no less than 8 seconds, which I've observed is pretty close to the natural period). I've also run pressure controllers without any integration, but find a bit usually gives better performance.

Although I'd prefer to keep the prop band wider there are trade-offs:

1. These machines don't have winders, so there is a benefit in obtaining a usable sheet width ASAP on start-up. A tighter P band gives the response necessary to allow this to occur. It would really be nice if the extruson control system were able to automatically sense the difference between a start-up and normal run condition, and use different proportioning terms for each
   
2. If the instability is caused by too much prop gain then widening the bandwidth is curative. However, if the underlying instability is "in the machine", then opening up the band may not allow fast enough response to these disturbances. In our application, if sheet width is lost then the ramifications downstream (sheet falling into former - crash, boom, burn) pose a bigger problem than touchier response in the extruder.
   

I've 'zoomed' into drive and pressuretrends for a typical instability episode. Suction pressure ranges between about 1350 to 1450 PSI (1400 SP), drops suddenly (within 10 seconds) to about 860 PSI, swings back to 1670 PSI 10 seconds after that, then resumes relatively calm operation.

The drive trend RPM is relatively stable (+/- 2 RPM) and only spikes as the melt pump controller responds to the sudden loss in suction pressure that appears to have been caused by the loss in screw load. Screw speed stayed high until suction pressure was restored.

I agree wholeheartedly that swings in screw RPM (by changing the rate of shear) affect screw stage one temperatures. I believe this interaction constitutes a complex feedback element since shear rate change may affect melt viscosity (and hence pumping rate) non-linearly. Less shear means less frictional heat produced, and vica versa. The temperature controller will continue to provide heat/cool output out of phase with actual conditions for some time (due to time lags discussed earlier), so it would appear less dramatic screw RPM changes lead to better thermal stability.

A greater instability in screw load is observed that doesn't track only with screw speed changes (which would be the case if they were only accel/decel-related). I'm thinking this points more to discontinuities in melt flow in the first stage (caused perhaps by bed breakup?).

Have a relaxing labor day holiday,

Bob

posted September 01, 2002 10:50 PM               

We'll be putting both resin and reclaim material streams through the crystallizer.

Thanks for the reference to Erema - this approach sounds great (less heat = less material degradation = less cost), but it's too late now - we begin installation tomorrow.

quote:

---

...just a dumb ass sparkie who's spent too much time near extruders...

---

Is a "sparkie" another acronym for an electron-pusher? If it is, then we're in the same boat. My training is as an electronics tech, and my career (or perhaps better put, "sick obsession") is industrial maintenance.

posted September 01, 2002 11:40 PM               

Yeah, sparkie is Kiwi speak for an electrician- did my time as an armature winder. Then worked as a service tech for an instrumentation company. Extrusion is my obsession, along with sailing.

It will be interesting to see what happens when you start running dried Apet.

posted September 02, 2002 10:42 PM               

First I would like to say it is refreshing to find someone who actually tries to understand their extrusion process. Most approach it as an art form, which can't be understood or fully controlled. It truly is possible to control and master an extrusion process through science and understanding! I hope your process becomes a case study in how this is possible.

You seem to have a very good grasp of the water cooling process. I have also seen a lot of solenoid failures. If you find the cure, please share it.

One variation of the water cooling process involves cutting spiral grooves on the outside of the barrel, and then pounding copper pipe into the grooves. I think you can see how this might improve cooling control. I don't see this done very often however.

A large diameter extruder has limited capability to control melt temperature. I think what should be achieved is a stable environment for the melting and pumping processes. Once that has been achieved, then pressure surging issues can be attributed to other causes. Sometimes you have to set the barrel zone to a temperature that the process is "Happy" with, as opposed to a temperature you might want to achieve. This in itself can markedly improve stability.

With regards to pressure control, this is very analogous to gravimetric feeding. If the feeder can't consistently deliver material at a set screw RPM within say +/- 3% short term and +/- 10% long term then you might as well stop and redesign the feeder. No PID control can make the proper corrections, and it will always be wrong chasing the last unstable result. I have seen failed attempts to control a feeder in this condition time and again.

Your extruder is exactly like the feeder and the suction pressure gauges its performance. I would therefore tend to screw designs that provide the most consistency in material delivery. The entire feed rate is determined by the fist stage of the screw. Pressure consumers (Maddox, Egan) and solid bed formers (barrier sections) will tend to help improve the output consistency. They will also tend to reduce it.

Some lessons have been learned in the "Direct to Sheet" lines using twin screw extruders and gear pumps. Since a gravimetric feeder controls extruder screw output rate starve feeding the screw, very stable feeding is required. The feeder is started up slower than required, and very gradually (High P) brought up to the required suction pressure. Real help can be had in the form of screw design in the second stage of the screw. Since twin screws are very efficient at developing pressure, minor excesses in feeding cause huge increases in suction pressure. Additionally the volume held by the discharge section was very small (at least before something bad happens), so minor reductions in feed rate would cause dramatic reductions in suction pressure. This problem could not be cured at the feeder because of a multitude of issues. The cure came in screw design. The discharge section basically had holes cut in the flights so that it would be less efficient. This allowed the discharge section to hold more material to act as a surge tank. The larger the surge tank, the slower the required feeder correction, and the more forgiving the process would be to momentary rate fluctuations. In fact you can build several surge tanks into a screw design to help stabilize the process.

Moral of the story:

If you have a shallow second stage on your screw, as would be a typical screw design, you have an efficient screw and a small surge tank. You may suffer sudden changes in pressure over the short term. An ideal design would be a long, deep cut discharge section. This would give you a larger surge tank. The surge tank would allow for more gradual changes in screw RPM to adjust to changes in suction pressure. More gradual changes lead to more temperature stability, and more temperature stability leads to less process instability and so on.

In fact I feel that the ideal size of the surge tank can be calculated based on the design of the first stage and how the process is operated.

Just some thoughts

Tom C

posted September 11, 2002 03:18 AM               

It's been fun swapping thoughts back and forth on this topic.

I like the idea of pushing pipe right into an externally grooved barrel - it sounds effective, but difficult to pull off. I'd imagine the grooving has to be done so it's shape conforms to the tubing OD, and go deep enough into the barrel so tubing OD ends up flush with the OD of the barrel (but not so deep as to decrease the burst strength). It would get rid of a large time lag component, and sounds like thermal resistance would have a net decrease as well. Hmmm ...

quote:

---

This allowed the discharge section to hold more material to act as a surge tank. The larger the surge tank, the slower the required feeder correction, and the more forgiving the process would be to momentary rate fluctuations. In fact you can build several surge tanks into a screw design to help stabilize the process.

---

Now, I've never thought about an extruder screw in terms of containing a surge tank (or multiple tanks), but this makes good sense. Was asking around today to see if we had drawings on this screw ... we don't, and looks like if I really want to know anything about it it'll be after a screw pull, and a boatload of measurements. I also like the analogy between gravimetric feeder and a screw delivering melt to a pump.

What you say about science vs. art (and, IMHO, art sometimes tinged in magic both black and white) resonates. It took me a while to learn that approaching tasks in a more-or-less scientific manner, although difficult and time consuming at the onset, usually ends up being the shorter path to success. Or, at least, helps to point out paths that don't go there ...

I'll let you know if we end up doing something different with barrel cooling, although we've been talking about it for a couple of years now, and simply haven't had the 'elective' time to spend in this pursuit.

posted October 14, 2004 09:09 AM               

A few questions for the board:
1) Is not better to have shorter cylce time for cooling or heating of the barrels? I have been told in the case of heating is best to have short cycles such as 5-10 seconds. The output of the PID will be a TPO. If the PID calls for 100%, then the output will stay on for 5 seconds. If the PID says 50%, then it will stay on for 2.5 seconds and off for 2.5 seconds. How about cooling cycle time?

2)If my cooling cycle time let's say 20 seconds and the controller is calling for 50% output, does it mean that the controller should stay on for 10 seconds and off for 10 seconds? or I have to do something else with my output before going out to the solinod valve because of the non-linearity of water cooling. For example, in this case rather 10 seconds on, should I change it to x seconds based on non-linearity calculations.

3)I am am planning on using a AB SLC PID to control the cooling output. The plan is not to start the cooling unless the heating output is all the way down to zero, then cooling takes over. This way it will also conserve energy. What kids of parameters you guys recommend for cooling as far as P,I, D is concerened?

posted October 14, 2004 11:52 AM               

Allow the PID loops to control with minimal interferance, like shutting off the heat too soon. Otherwise you will makes the PID loop go crazy.

Tuning is simple, but different systems seem to use different values so it is hard to provide numbers. Tune while running if possible with material that can tolerate large temperature swings. At first set P heating to a mid value and turn heat I and heat D off. Turn the cooling off entirely. Try a P value for heating. Increase the set temperature by 50F. Study the rate of change of the temperature and the overshoot. If the rate of change is too slow, double P. If the overshoot is more than 10F, half P. Change P unitl the system the heating is reasonably fast, but does not overshoot much.

Let the zone run for a while. The temperature should control to within 5-10F. Now is the time to incorporate I. Try a value for I. If the temperature varies around the setpoint, reduce I. If there is too slow a correction increase I. The D value is often I/6 to I/10.

Tuning of cooling is much the same. Make sure the cooling water flow to the zone is low before tuning. Tuning and zone performance will tell you if more water volume is needed.

Tom C

posted October 18, 2004 01:15 PM               

I have also found that adjusting the time base for cooling (the period of time between consecutive openings of the solinoid) to be very helpful. If you can set the time base to something like 20 seconds, the maximum on time to 2 seconds, proportionatly control withing the 2 seconds, and set the needle valve to a fairly high flow, you can keep the control in the "steam flashing" zone. This will give you much more control by not using a liquid water flow, which has different cooling characterisitcs.

posted October 18, 2004 03:34 PM               

Sorry about the weird formatting. This was the only way I could figure out how to geerate columns.

The reason this works is that the long off time allows the water channel tube to reheat to a temperature where the next shot of water will be flash steamed. If the water is on too long, or off too short of a time then the water tubes will cool down too much and will no longer flash the water.

Tom C

posted October 19, 2004 08:29 AM               

What algorithm you use depends on the process and the degree of control required. Generally extrusion is supposed to be a steady process, so the derivative term could casue more problems than help. I think you need to set up a proportion band and use integration. Integration of the error shifts the proportional band, which allows greater control. This would seem very important for water cooling. If your extrusion process is generating heat in a zone, PI control should be able to find a cooling output amount that will pull away the heat being generated, and provide a stable temperature.

posted October 19, 2004 10:42 PM               

You stated the followings:

"Heating and cooling PID loops should be run as separate and independent control loops. In other words the heating and cooling loops don't know what each other are doing. Only one loop should control at a time however. In order to determine which loop controls, a logic statement should evaluate whether the current loop has zero ouptut. If the other loop is not zero, and the temperature has crossed the setpoint plus a deadband (+/- 5F) then switch to the other loop. When the switchover is done, the control logic should reset the P and I control values of the newly activated loop. Use the D value for a starting control output point and activate the loop. This way a rapidly changing temperature will be handeled quickly on the switchover, but a stable slow drift will see little correction. Allow the loop to run from there. "

Let me describe what I have and tell me if it is different than your statement.

I have two separate PID controls (cooling & heating)with different K,I,and D parameters. If the extruder temperature is greater than the setpoint and the heating controller output is zero, then the cooling PID takes over. The key here is that I wait until the heating PID is done everything it can to reduce the heat and if that is not enough then the cooling controller will take over.

Is this the same as what you were saying?
Just a quick point, my valves for the cooling are Asco type and it is not needle type, Would that make any difference as far as the percentage output you taked about in your last reply?

Sam

posted October 20, 2004 07:10 AM               

There should be a neddle valve in line with the solinoid valve.

Using the 20 second time base for cooling you can set a fairly high water flow rate.

posted October 20, 2004 07:06 PM               

In regard to the neddle valve, I was planning to keep it wide open and let the Asco do the control. It is easier to tell the extruder operators not to play with the neddle valve, just leave it alone.
Do you see a reason not to have the neddle valve wide open? I am going to assume with the quick output pulses, I should have a good control.

posted October 21, 2004 12:35 PM               

Since you have been helpful, one more question:

I am not sure which control scheme is better.
In my previous posts, I indicated that if the temperature is greater than the setpoint, the cooling loop will not do anything until the heating output is all the way down to zero. Two ways of doing the cooling.

1) Since I have a separate loop for cooling and soon as the temperature is greater the setpoint, the loop will start calculating the percent output, but NO output will be going to Asco's beacuse the heating loop output is not zero yet.

2) Enable the cooling PID calculations when the heating loop output is zero and the temperature is still greater than the setpoint.

In the case #1, the cooling loop has been calculating an output, it is just waiting for heating output to go to zero. I am afraid in this case because of the Integral action the output may already winded up and may try to put too much water.

In the case #2, I am basically starting with zero output, because I will enable the cooling loop calculation as soon as the heating output is zero.

What are your thoughts on this one?

posted October 21, 2004 06:40 PM               

Much of contol technology hopes to control dynamic systems. Ideally an extruder should be at steady state. I have found that tuning for the steady state, and not worrying about the wild ups and downs of startups is better.

In cases where the resin is very sensitive to overheating, like PVC, then a more aggressive approach such as option 1 might be best. Sometimes when dealing with resins like PVC there is fine line between allwoing the system to overheat by being too conservatve with cooling, or torquing out the extruder by being too aggressive.

Put you control loop together, and hopfully you can plot the temperature and control outputs over time. A nice operator interface will let you change a lot of parameters as the line is running. The results of those changes can be seen. If you can't tune on the screen, then perhaps the algorithms need modification, or the heating and cooling system need adjustment.

Don't expect perfect temperature control with extruders, I've rarely seen it.

Tom C

posted December 07, 2004 02:30 PM               

If you can not acheive a constant offset by this method, then you need to see what else is upsetting the extruder's stability.

The extruder's stability can be observed through the motor amps and head pressure. Please report how much variabilty you see in those values, or if you can get stable temperatures.

Tom C

www. ExtrusionTech.com

posted December 10, 2004 12:03 AM               

Here I was pondering only the load upset question, and not what to do for setpoint changes. A thumbnail sketch of the idea was to define a 'primary' cycle time (such as 20 seconds), and a 'sliding inner loop' cycle time that would range from as short as 2 seconds, and as long at the primary cycle time.

Cooling PID would control how long the cooling valve was opened within that 'inner loop' duration (for instance, if the inner loop time was 2 seconds, and 50% cooling was called, then the valve would be on 1 second, and off for the remainder of the 'primary cycle time' duration (at 50% cooling, for 19 seconds).

If enough cooling isn't possible with the 2 second on time @ 100% cooling duration then this timer 'slides' to a higher setpoint (4, 6 ... up to the 'primary' cycle time) until controllable cooling is obtained.

Not discussed was what to do when cooling requirements change due to setpoint changes. One possibility would be to no nothing if the setpoint change resulted in a higher temperature, but temporarily disable this proposed cooling modification until temperature was again within x degrees of setpoint when a decreased setpoint is desired.

This would be analogous to what Eurotherm does with integration by using high and low cutback terms (here, the integration term is ignored to prevent wind-up when MV is too far from SP).

One thing you might find useful is to (at least for experimentation purposes) provide a manual cooling mode, and manually set cooling on time rather than letting PID determine it. You can opt to leave the heating side alone ... let it remain under PID control ... or put that in manual as well with a 0% setpoint.

This way you can stand at the machine, and by trial and error find the cooling on time that is just long enough to give enough cooling under that particular set of operating conditions. This info is useful when troubleshooting the loop once it's put back into auto ...

Bob

posted July 24, 2005 03:44 PM            

In refer to cooling system topic for extruder barrel.
Current Condition
(i) Extruder mixing zone always display overheat temperature by 20-80oF with respec to preset temperature of 410oF. It also depends on percentage amount of HDPE and reclaim presence in the layer. (Higher the worst)
(ii) I discovered only cooling fan use as our cooling system. Besides, our technical has removed at least 3 heat band. Though, it only help lower 3-6oF.
(iii) the screw yield recorded at 18-21.

Actions we taken:
(i) Decrease output from 1300lb/hr to 1000lb/hr.
(ii) clean cooling fan from dirt attached.
(iii) supply cooling air directly into barrel zone.
(iv) Change screen pad periodically when melt pressure recorded high.

We are concerned about film quality / physical properties degradation due to low output and overheat (overcook).