My first AutoCascade -- building has begun

[n00b 0f l337]

I don't see it on the diagram, just a watercooled condenser?

[mytekcontrols]

I don't see it on the diagram, just a watercooled condenser?

Opps! You are correct. I mistook the Primary Condenser as being the Auxiliary Condenser. So apparently this AutoC design only has 2 Cascade Condensers, and nothing else. Although this will work, it isn't exactly the most efficient method, and fails to take advantage of the additional cooling that would be available with an Aux Condenser proceeding the 1st phase separator. So in essence the 1st phase sep will run close to ambient temperatures, instead of -25 to -35 C with the addition of an Aux Condenser.

Here's a diagram showing the best orientation for refrigeration applications using flat plate HX's...

HXflow2.jpg

Condensing should be traveling downward, and evaporating upward. In the case of the water condenser, the water would be traveling upward.

[n00b 0f l337]

Correct.
However, I see in autocascades that some of the warmer refrigerants may make it to an evaporator. In which case they may pool and remain liquid or even freeze. A coil evaporator solves the issue by evaporating downward in a linear fashion, while a plate HX provides ample room for pooling in the volume at divider tubes.
If I can read the tiny font though, his first phase sep is sub zero. Is there a HX not being labelled?

[mytekcontrols]

Correct.
However, I see in autocascades that some of the warmer refrigerants may make it to an evaporator. In which case they may pool and remain liquid or even freeze. A coil evaporator solves the issue by evaporating downward in a linear fashion, while a plate HX provides ample room for pooling in the volume at divider tubes.
If I can read the tiny font though, his first phase sep is sub zero. Is there a HX not being labelled?

If you had an iPad you could zoom up on it, else save the image to your computer and then use a photo viewer to do the same. Anyway I see it as +28.4 C (KTIT 102).

[n00b 0f l337]

I'm looking at KTIT 108.

[mytekcontrols]

I'm looking at KTIT 108.

That's the expansion point of the 1st phase separator's cap tube, which should not be confused with the actual temperature of the phase separator itself. So basically what we have here is a #1 phase separator running at 28 C (essentially like liquid line temperature on an single stage unit), and evaporating downstream at -9 C. Not very cold as compared with most of the single stage systems you guys build. In fact it happens to be nearly the same as the boiling point of Isobutane near 0 psig. But with a running suction pressure of about 33 psig, it suggests that some of the lower boiling component(s) are also at play.

[n00b 0f l337]

Ah gotcha. Then you also think a HX before this would help tremendously? For all its worth it appears he's simply built a beautiful cascade and using one large enough compressor.

[mytekcontrols]

Ah gotcha. Then you also think a HX before this would help tremendously? For all its worth it appears he's simply built a beautiful cascade and using one large enough compressor.

I think Kevin and his team have done a great job

And it is especially nice that he has been willing to share the progress of this build in such amazing detail. However with all that said, it like all prototypes, probably still has some more optimizing in it's future. With that in mind, and to answer your question, yes i think it would be even better with an auxiliary condenser (reference 3-stage w/sub-cooler flat plate design below).

FP-HX_Stack.png

(Blue arrows = Suction, Red arrows = Discharge)

Edit: The flat plate HX's are all 40 platters.

This drawing is incomplete (no cap tubes or evaporator connection shown), but it does show a compact way to build a flat plate stack while still adhering to the vertical orientation requirements of the plates. You'll also notice that the Auxiliary Condenser is much larger than the plates used for the Cascades. This drawing was based on a real working production system, and early on in it's development cycle it used identical plates for all the stages. It was later found that additional work was possible by up-sizing the Aux Condenser, and better performance was gained by doing so.

BTW this particular system handles a load of 3500 watts while maintaining -100 C or better across the evaporator.

So based on this and numerous other designs that I have tinkered with (including CryoBUG), I would definitely have to say a big fat yes to adding an Aux Condenser.

[Kevin Hotton]

Hi Michael and nOOb,


The reason I oriented the cascade condenser nearly flat (they have a slight tilt) is that the quality at the outlets are two-phase flows (vapor and liquid) with quality about 0.5 on the high pressure side and nearer 0.9 on the low pressure side. If I mounted them vertically and there was insufficient flow velocity to clear all the liquid, I could pool liquid if mounted vertically and upset refrigerant composition (I discussed this in an earlier post). The water cooled condenser has full vapor inlet so horizontal mounting will not affect HX flow distribution (which will be uniform across all plates). The final evaporator was mounted vertically to assure full saturated vapor outlet condition.

We did add a flow distribution “comb” that inserted into the HX refrigerant inlet port. This tube extended to the full depth of the HX and had a line of many small holes to distribute the refrigerant evenly over all flow passages in the 40-plate HX. Before we did this the overall heat-transfer coefficient of this HX was terrible at 40 W/m2-K, now it is only “bad” at about 80 W/m2-K. The first cascade condenser (warmer one) has an overall heat-transfer coefficient of 175 W/m2-K and the colder one is 111 W/m2-K. Both these values are about half of what I expected and is the primary reason for the shortfall in cooling duty.

Below is a screenshot of today's run (we ran steady-state for nearly 2-hours). The water-cooled condenser flow meter is not working well so the cooling duty is not correct; the balance of the data is correct (I think). The Hysys simulation has proven to be quite accurate (all measured temperatures are within 6.5C of prediction -- compressor outlet temperature is the exception). The current cycle still has too high of a discharge pressure. I have shortened the captubes and the suction pressure is now higher. I am refining the Hysys model (with the poor HX performance included) and have a new cycle that is predicted to have the same cooling duty, but a lower discharge pressure (235-psig instead of 274-psig). This will improve the cycle's COP (coefficient of performance; same cooling but lower power consumption). You can see we are matching nicely the -30C to -60C Methanol cooling requirement.

Finally, the reason I did not add an auxilary condenser is that by Hysys model simulation what I discovered is this: Adding one consumes useful cooling duty (in my case). You see I am chilling methanol from -30C to -60C. I can take advantage of temperature "glide" which I clearly have in my evaporator (-73C to -44C). What an auxiliary condenser does is use the cooling duty that I am using from about quality 0.8 to 1.0 (were glide is most pronounced) and instead cools the first phase-separator so that a higher-vapor-pressure refrigerant blend can be used -- with the final result being a slightly colder evaporator inlet and an outlet with less temperature glide, but no more cooling duty. For my application (unless I am mistaken and made a mistake) an auxilary condenser does not help.

[n00b 0f l337]

I'm not sure that I see that point of taking cooling duty.
Effectively the aux condenser only uses "post-process" refrigerant. That is, refrigerant that is otherwise wasted and will simply lend to compressor temperatures. However, this is compensated for normally by lower operating pressures.
Furthermore, since you are just running a two refrigerant blend, I still see the extra phase separation as less than useful, versus adding a third component to further drive down temperatures or increase duty.

What is the name of the software? I'm interested in researching it.

[mytekcontrols]

Hi Kevin,

Very interesting on the comparison tests with and without the "comb" distribution tube. I would judge by the bad performance without the comb, that the horizontal orientation of the HX was causing you problems. And although the comb has made things better, you'll never know for sure if it is as good as having the HX's vertical. I understand your reluctance to make the changes required to do this, but this is after all a prototype, and prototypes often require tearing things apart and rebuilding (that's why I've recently learned to make mine strictly as bread boards, so that I don't care about keeping it pretty). I can tell by the way you constructed your prototype, that a lot of care went into making everything look good (including a pretty blue painted frame). But as you said you seem to be achieving your goal, and it does line up with your simulations, so who am I to tell you otherwise.

But because I can't resist, this is what I would do as individual steps to insure that you are getting everything that you can from your prototype before building the final product.

1) Try HX's in a vertical configuration.
2) Try adding an Auxiliary Condenser before phase sep #1.
3) Replace liquid dryers with cap tube bullet strainers
4) Cut apart what you're using as a phase sep and determine if and how much liquid backs up before feeding out.
5) Identify and Minimize any source of liquid hang-up, and let gravity work for you. This is especially important in the colder stages.

And just from my personal experience, don't assume that what you have built will continue to do what it did in the lab when it is being used in it's final application. Test it, then test it some more, and then beta test it in the field for a few months, before you say it's done. You wont believe how many times we had units fail when they went out to the customers and encountered something that had not been anticipated. AutoC's are trickier than they seem.

Please take all this as me trying to help, and not in any way trying to diminish what you guys have accomplished. And for someone that has never built an AutoC before (I think I got that right?), you certainly have caught on quite well, and are already far ahead of the learning curve as compared to many others trying to get these Damm things to work.

[Kevin Hotton]

Hi nOOb,

Here is the Aspen Hysys model (that's the chemical engineering software I use) of my AutoCascade with the added auxilary condenser. You can see that the temperature glide is gone (-76C evap. IN, -74C evap. OUT). However, the cooling duty is no greater (I am modeling the same scroll compressor I am using). The cycle's COP is also nearly unchanged -- no better. Also, the size of the second cascade condenser is about twice as large because of the reduced temperature differences in this cycle. I admit the auxilary condenser is only cooling from 28C to 18.5C, but that is all that is necessary if a minimum temperature of -75C is what is desired. I am definitely NOT an expert AutoCascade designer (this is my first one). But software modeling has given my a lot of insight into the interplay between the various HX's in an AutoCascade. You really have to finesse a design to match all constraints (getting sufficient delta temperatures across ALL HX's is the most difficult).

note: HX = heat-exchanger

I may be mistaken, but based on the modeling I've done, an auxilary condenser is not an improvement for my application.

Kevin

[Kevin Hotton]

"Very interesting on the comparison tests with and without the "comb" distribution tube. I would judge by the bad performance without the comb, that the horizontal orientation of the HX was causing you problems. And although the comb has made things better, you'll never know for sure if it is as good as having the HX's vertical."

NO Michael, the HX with the "comb" is VERTICAL. It is the final evaporator. The horizontal one's are out-performing it.

Also,
1) the filter/driers have been removed and bullet strainer in their place.
2) flow is now either vertically down into captube (final one) or flat horizontal in (first two captubes).
3) phase-separators (really oil-separators) will not trap liquid. We are NOT using the normal float-activated oil outlet, but a separate drain outlet right out the bottom.


Kevin

[n00b 0f l337]

Interesting. A vertical PHX wouldn't appear to need a comb though. I think it might be assisting in reducing liquid logging of isobutane, but not solving the issue.
Only thing really to try would be to swap it for a coil style HX, and see if that improves the situation.

I see that you've done the modelling, but ultimately with modelling I often see artificial constraints, and some things remain unaccounted for. A common factor seen in autocascades is the lower temperature refrigerant. Lower so then what you are using, and normally there would be a third refrigerant employed in your system. I think your refrigerant selection maybe showing its limitations, as well as the evaporator issue. Additionally, in all models I have seen, the ultimate temperature after the auxillary condenser is FAR lower then a 3 C drop. Even at similar load configurations to your system. More often then not, you see 20-30C differences.

Interestingly in many of Myteks threads, you notice that this temperature is much lower.
I think your modelling may not account for this properly, or the charge you chose for the modelling does not take this into account either.

You've mentioned multiple times also that you've charged to liquid in the sight glasses. This seems in grave error as with most capillary systems, a liquid seal is important, however a flood of liquid is not needed. Refrigerant will condenser in the capillary lines, or will provide cooling via an adiabatic process cooling. This may result in that you have overcharged.

Just some thoughts I see after looking over the data.

[Kevin Hotton]

Hi nOOb,

These are my thoughts on our evaporator "before and after "COMB" modification". As mentioned in an earlier post, the liquid fraction flow rate into the evaporator is only about 1-oz every 5-seconds. So this small amount of liquid can easily pool in the bottom of the plate HX, while the much more volumous vapor fraction shots up the plates (providing little cooling because it has only sensible heat to give). After a time the liquid accumulating in the bottom of the HX will be blown toward the back and begin to enter a few of the flow passages. Of couse on the opposite side methanol flows but, now the case is too much liquid refrigerant relative to the amount of methanol to cool. So the end result is: most of the methanol is only being cooled very little by vapor and a small amount of methanol is being cool greatly. However, the mixed methanol overalll is not being efficiently cooled at all (i.e. bad overall heat-transfer coefficient for the evaporator).

The "comb" remedied this. It forces a more uniform distribution of refrigerant liquid / vapor across the entire HX with the result being seen as a doubling of the HX's overall heat-transfer coefficient (this allow the requried delta T to be cut in half for the same cooling duty). This "comb" is not a new idea of mine -- something similar is in the GEA flat-plate HX's (the ones that are specifically sold as refrigerant evaporators).l

Also, AutoCascades with two phase-separators do not necessarily need three refrigerants. Michael discussed this several times in his "morphed AC" post.

Further, I don't think an auxilary condenser helps in my application; I am sure that it does if I am after a -90C (or below) evaporator temperature. I also tried cycles with an auxilary condenser and a single phase-separator, but found none that were as good as the system that I have built.

And Finally, I can't claim infallibility for Aspen Hysys modelling, but for what I have experienced thus far, I am amazed at its accuracy. It really seems to capture all the thermodynamic nuances of an AutoCascade cycle.

Thank you all, for all the comments and advice given,

Kevin

Kevin

[mytekcontrols]

"Very interesting on the comparison tests with and without the "comb" distribution tube. I would judge by the bad performance without the comb, that the horizontal orientation of the HX was causing you problems. And although the comb has made things better, you'll never know for sure if it is as good as having the HX's vertical."

NO Michael, the HX with the "comb" is VERTICAL. It is the final evaporator. The horizontal one's are out-performing it.

Also,
1) the filter/driers have been removed and bullet strainer in their place.
2) flow is now either vertically down into captube (final one) or flat horizontal in (first two captubes).
3) phase-separators (really oil-separators) will not trap liquid. We are NOT using the normal float-activated oil outlet, but a separate drain outlet right out the bottom.


Kevin

Oops sorry I missed that being the vertically positioned evap that had the comb. So I guess that doesn't do a very good job of supporting my recommendation on vertical versus horizontal for all the HX's. However I still believe vertical is preferred for refrigeration use, and there are many supporting views on this by various flat plate manufacturers. So it would still be a viable option to try at some point.

Glad to hear you made some of the changes that were recommended earlier.

On the oil seps, yes I know they are non-floated, and really shouldn't hold up much liquid (much being the key word). But I know on a relatively small oil sep I was using, it actually held up a relatively large amount proportional to its size. This probably wouldn't be an issue in the 1st phase sep where there should be plenty of liquid refrigerant, but could be a problem in the 2nd. Just something to check to be sure.

As Adam pointed out, an Auxiliary Condenser normally makes a major difference in the temperature of the outlet versus the inlet. In most cases it will take it down at least 35-40 degrees even at full load. And I have never seen it be a detriment to the system's performance, but always an improvement. Perhaps the simulation is flawed, no disrespect to you, but this would not surprise me. With all the possible variables at play, I really don't see how the simulation could be perfect the first time around. It would be nice, but not likely.

But if this software can accurately predict every possibility, I was wondering if we could try a simulation of the next Lottery?

Bottom line is that I have to say the software does appear to be working quite well for you.

[mytekcontrols]

One last thought before I drift off to sleep.

Kevin,

Have you tried running a simulation with only one phase separator preceded by an Auxiliary Condenser and followed by a single Cascade Condenser? This would be appropriately called a single-stage autocascade. For the temperature you are after, which is not all that cold, it would seem like a more efficient option. And would probably work quite well with only two refrigerants in the blend.

Since you are using oil seps, having one less stage of phase separation shouldn't cause any oil problems, even though the one phase sep will be running very cold.

[n00b 0f l337]

You say liquid and gas into the evaporator at the end. You really should have no gas entering, but I think I know what you mean. Your low stage refrigerant enters and boils, then runs up the plates. This is inherently the issue with using a plate heat exchanger as the final evaporator. Much of the heat removal is from the actual conversion of liquid to gas, and gas will flow more quickly and transfer less with the walls. The evaporator is cold enough that the isobutane at the bottom may never actually boil off in operation. It will simply pool and provide a liquid mass of cold refrigerant that transfers against the area its near.

In other points of the system, butane should not pool. I believe your lack of efficiency at the evaporator is due to this, and you would be best suited with a series evaporator (like the coil-in-coil style) versus the parallel style you have in a PHX. This also serves to deliver the butane back through the evaporator to the previous heat exchangers, increasing efficiency.

A possible methodology in testing pooling, involves running the machine, and then turning off the load and machine. From here you can watch the evaporator temperature as it slowly returns to ambient. If there is pooling, after a time the sound of boiling refrigerant will cease as all the ethylene (I believe it was ethylene), has evaporated. After this silent time though, the temperature/pressure will hit a point where butane should return to a gas if your static pressure is low enough for it to do so. This will bring back the boiling noise. However, your pressures maybe high enough, and your evaporator temperature may never reach a temperature that will allow the butane to boil.

It could simply be accumulating and when the machine starts up each time, returning the liquid volume out of the heat exchanger as pressure rises high enough to deliver it out. Or it may stay suck.

I don't think the comb is doing you much good. If anything, depending on its size, it may have just displaced trapped butane. Or it is in a way working as described.

Just my theory. And with the butane pooling, you would see the warmer return temperatures at the lower HX's, and would be missing the improvement of an aux cooler.

I understand also your 2nd phase separator is helping to increase separation and purify the stream. It may even help to keep the butane out. But without it, and an aux condenser, as Mytek suggested above, I believe you could see the same performance or better as the design is simply more efficient as seen by its use for the past decades by autocascade technicians at Polycold and competitors.



Further edit
As seen in your last system picture, I notice a distinctly larger delta at the low stage exchanger KTIT110/111.
Similarly, a warmer temperature significantly at KTIT105. However, in the high stage exchanger, the delta of about 13C is still there. So KTIT106 is ALSO warmer.
But at the low stage exchanger, your delta is 20C.

Working on what to make of it, but you clearly have worse transfer at the low stage condenser then you had before.

The final evaporator has a 13C delta, this seems to fit the plate heat exchanger working actually, just fine.
So I wonder why you feel it isn't being efficient?
It also stands to reason that transfer is dependent on a difference in transfer. Achieving a lower temperature using another low stage refrigerant, would also improve the heat transfer to the methanol. It doesn't mean you will get the methanol colder, although it might and then allow you to cycle the either the system or the system chained to the methanol loop. In this situation I'd imagine you are filling a resevoir with the target temperature methanol, and then orchestrating it's flow to the subsystems that need cooling.
If the loop is indeed too cold, cycle it. If not, improved transfer at higher loads as I believe you have expressed you wish to achieve more cooling duty.

[mytekcontrols]

A simple solution for the system running too cold, would be to put an EPR valve in the compressor suction line (as close to the compressor as possible). By adjusting the setting of the valve, you in essence have temperature regulation via pressure regulation. It works quite well on Polycold 552 and 662/672 units, and it does so without the need for any electronics, or cryogenic valves. The Sporlan ORIT series would do the job nicely (specifically ORIT-6 or ORIT-10).

This is actually a better way to accommodate a specific temperature requirement instead of trying to "tune" the charge to comply, which can't respond to unanticipated load changes or water temperature variation. Always better to have reserve capacity and/or colder temperatures then needed, with regulation to hold it to where you want it to be.

ORIT_valve.jpg
ORIT-6-0/50 available at Amazon for a good price ($40-$50).

[n00b 0f l337]

How does that valve function? Simply controls the suction pressure?

[mytekcontrols]

How does that valve function? Simply controls the suction pressure?

It's a pressure regulator. By looking at the pressure on the upstream tap, and adjusting the set screw, you can set the pressure you wish the upstream side to run at. The valve will then automatically open and close in order to maintain this pressure setting. And by increasing the pressure of the evaporating side, or suction side of the AutoC's HX stack, you will also increase the temperature that the various refrigerants will boil off at. Thereby giving you automatic temperature control.

Trust me, it works quite well.

[Kevin Hotton]

Hi nOOb and Michael,

Thank you for all the great advice and insights. I just tried the GEA HX sizing software (free download) for sizing my evaporator and it says, “channel velocity is too low” and won’t suggest a HX. The problem seems to be the large methanol temperature change of 30C which only needs a flow rate of 2-Liters/minute. However, it would not be easy to make a tube-in-tube HX. The duty I am targeting is 1800W and to acheive this would require three (in parallel) two 3/16 inch tubes within a 1/2 inch tube coil each 18-feet long.

We ran our AutoCascade today, and performance was similar to our last run. We tried lengthening and shortening the final captube by one foot each way and both were worse. So the best has been: all captubes are 0.064 inch ID, the first one is 7-feet long, the second 8-feet long, and the final one is 6-feet long. The reason we are not getting optimal cycle efficiency is entirely the HX's. If the HX's were working bettter, the temperature differances between the counter-flowing streams would be less and the overall cycle efficiency would be higher. I was expecting to have about 10C delta across all HX's, but to transfer the heat it is taking over 15C.

I have revisited the cycle designs with a single phase-separator and an auxiliary condenser. With my new knowledge, I have derived a new cycle that is looking good. I do seem to need three refrigerants: iso-pentane, propene, and ethylene (in 36%, 26%, 38% weight fractions). The cycle efficiency appears to be better than the two-phase-separator design we have built. Below is an attached simulation diagram. Again, the greatest difficulty is getting sufficient temperature differance across the HX's, without the "help" of this Hysys program, I don't think I could have derived this solution. Even though this cycle has a higher COP (coefficienct of performance) the total duty is less than my original goal of 1800W. However, with the HX difficulties I have been having (limiting duty to about 1200W) this cycle's +1300W duty is an improvement.


Kevin

[Kevin Hotton]

Here's a screenshot from today's run. Again we ran steady-state for over 2-hours. As you can see the temperature agreement with the Hysys model is very good (all temps within 5C of design-point).

Kevin, Matt, and Joe

[Hondacity]

can i see a picture of your loader? I'm super impressed with the 5C in-spec results.

[Kevin Hotton]

can i see a picture of your loader? I'm super impressed with the 5C in-spec results.

I am not sure what you mean by "loader"?

Kevin