My first AutoCascade -- building has begun

[mytekcontrols]

I believe the very large helical oil-separators I am using for phase-separators is requiring a large refrigerant inventory (if I did this over I would use smaller phase-separators).

This would only be true if there were significant liquid hold-up. A while back I mentioned about checking how much internal liquid volume it takes before it is seen exiting the separator. Did you ever perform such a test?

Otherwise having a large phase separator gas volume should not be an issue, and if anything, aids in separating the liquid from the gas.

I also believe removing the suction accumulator was a positive step.

I find it strange that the suction accumulator would be a problem. On rotaries it never seems to be an issue.

I went back and reviewed data along with my captube calculations and convinced myself that I need to make all the captubes longer.

Looking at your diagram, I see you had a 32.5 psig suction versus a 289.4 psig discharge pressure. This ratio doesn't look bad to me, especially considering the load. So although lengthening the cap tubes will bring the suction pressure down, theoretically allowing the refrigerants to evaporate colder, it probably will also mean less heat load capacity.

The remaining problem is all about the HX's we are using. Notice that we have increased the methanol flow to over 3.5 LPM to heip heat-transfer. The evaporator is cold -73C but the exit methanol is only -54C (poor temperature approach).

Once again looking at your diagram, your temperature split on the primary mixed refrigerant side of the evaporator is starting to look pretty big at 27 degrees C. A more typical temperature gradient under full load should be 10-15 degrees C. So this would suggest three possibilities...

1. Insufficient liquid feeding the evaporator
2. Excessive system pressure drop (hard to say without downstream pressure readings)
3. You need a bigger compressor

If the evaporator HX had poor heat transfer as you suggest, then I would expect to see the feed and return on the primary side to be very close in temperature (very little gradient). the fact that the temperature gradient is large, suggests that very good heat transfer is taking place (assuming that you are not simply running out of refrigerant). Personally I think you still need more low boiling refrigerant flowing in the system. And judging by the compressor discharge temperature of 108 C, you might need a tad bit more butane (you might find that mixing a bit of n-Butane with the iso-Butane would also help in this area). Large systems usually require either a high static balance pressure (225+ psig) or sufficiently sized expansion volume, in order to provide enough low boiling refrigerant when condensed.

[kang-China]

yes ,i gree mytek
your comp is not biger in your system

[Kevin Hotton]

Hi Michael,

I very much appreciate your expert advise.

This would only be true if there were significant liquid hold-up. A while back I mentioned about checking how much internal liquid volume it takes before it is seen exiting the separator. Did you ever perform such a test?

No I have not, but as I have written in an earlier post, we are exiting out of a drain plug in the very bottom (not the float activated one which we have capped off). The first captube begin flowing liquid within the first 30-second, the next in about one minute (and the last one feeding the evaporator begins flowing in about 2-3 minutes. I know I have a lot of thermal mass to cool down. What puzzles me is that starting at a charge of around 2.5-lbs, all the captubes are very close to completely full (sight-glass chamber is 90% or greater full). Yet adding one more pound of charge does improve cooling duty.

I find it strange that the suction accumulator would be a problem. On rotaries it never seems to be an issue.

Well in our case the suction accumulator was large and insulated. I am pretty sure that at a suction pressure of over 30-psig that liquid hold up of butane was occuring.

Looking at your diagram, I see you had a 32.5 psig suction versus a 289.4 psig discharge pressure. This ratio doesn't look bad to me, especially considering the load. So although lengthening the cap tubes will bring the suction pressure down, theoretically allowing the refrigerants to evaporate colder, it probably will also mean less heat load capacity.

No Michael, I have already lengthened the captubes and this test is the result. I am not planning on lengthening them further. This test used: 8-ft of 0.064" ID for the 1st captube, 11-ft of 0.064" for the 2nd, and 8-ft of 0.052" for the 3rd (at evaporator).

If the evaporator HX had poor heat transfer as you suggest, then I would expect to see the feed and return on the primary side to be very close in temperature (very little gradient). the fact that the temperature gradient is large, suggests that very good heat transfer is taking place

Yes heat-transfer of over 1600W is taking place, but under what conditions. Heat-transfer in a HX if calculated by the equation Q = U*A*LMTD. (Q is Heat in Watts, U is the Overall-heat-transfer-coefficient in W/m2-C, A is HX area in meters^2, and LMTD is the "Log-Mean-Temperature-Difference" this is the effective temperature "driving force" between the counter-flowing streams). So the large temperature difference between the counter-flowing streams means that the "U" is poor. A HX of a superior design will have a high value for "U" (this allows a closer approach in temperatures betwee the counter-flowing streams. Heat transfer is aided by turbulence (which is mainly caused by high flow velocities -- with its corresponding increased pressure-drop). This is why tube-in-tube designs work well (and seem to be favored by PolyCold).

Kevin

[mytekcontrols]

A HX of a superior design will have a high value for "U" (this allows a closer approach in temperatures betwee the counter-flowing streams. Heat transfer is aided by turbulence (which is mainly caused by high flow velocities -- with its corresponding increased pressure-drop). This is why tube-in-tube designs work well (and seem to be favored by PolyCold).

Very interesting on the comparison of the flat plate efficiency vs. the tube-in-tube HX, and something that I've begun to suspect recently. Perhaps Polycold will soon be in trouble, because they have abandoned the tube-in-tube HX in favor of the flat plate on there newest XC Chiller unit.

[ultralo1]

What puzzles me is that starting at a charge of around 2.5-lbs, all the captubes are very close to completely full (sight-glass chamber is 90% or greater full). Yet adding one more pound of charge does improve cooling duty.

A full sight glass only indicates that you have liquid up to the cap tube. It doesnt represent a full evap. By adding the additional pound of refrigerant, it increased the amount of refrigerant in the evap. Therefore you saw better cooling.

Kevin, this is not meant to be insulting or derogatory but a wordsmith i am not. You seem to be placing all your faith into what the model is telling you and not enough in the "art" of charging the system. There are variables that the model cannot take into account. That is where tweeking a charge becomes an art. use the model to get you close then add small amounts and record the data. Then do again and again until the data tells you that you have passed the optimal efficiency and are starting to go backwards. Soon you will start seeing patterns. These patterns will give you some insight as to what is actually happening in the system. So far I have seen you change both the system components and the charge, I have not seen you take a single system to its optimal performance as built. You are changing too many variables at once. There is a reason that the newest generation of 2 stage cascades have 10 thermocouples on them and the R&D models have over 25.

Britt

[n00b 0f l337]

Well said,
Data means nothing unless it is created with proper practice. Change one thing at a time.

[mytekcontrols]

A full sight glass only indicates that you have liquid up to the cap tube. It doesnt represent a full evap. By adding the additional pound of refrigerant, it increased the amount of refrigerant in the evap. Therefore you saw better cooling.

Kevin, this is not meant to be insulting or derogatory but a wordsmith i am not. You seem to be placing all your faith into what the model is telling you and not enough in the "art" of charging the system. There are variables that the model cannot take into account. That is where tweeking a charge becomes an art. use the model to get you close then add small amounts and record the data. Then do again and again until the data tells you that you have passed the optimal efficiency and are starting to go backwards. Soon you will start seeing patterns. These patterns will give you some insight as to what is actually happening in the system. So far I have seen you change both the system components and the charge, I have not seen you take a single system to its optimal performance as built. You are changing too many variables at once. There is a reason that the newest generation of 2 stage cascades have 10 thermocouples on them and the R&D models have over 25.

Britt

Kevin -- I also think that too much faith has been put into the simulation model. Perhaps after tweaking the data based on extensive tests with real hardware, the simulation can be trusted. But as you admitted being new to AutoC design, it would seem unlikely that the model is completely good at this point.

The methodology that Britt describes in his last paragraph (quoted above) is sound advice, and incorporates a lot of what I strive to do when I develop AutoC systems. Although sometimes I too get anxious, and change more then one thing at a time, but this usually bites me in the end.

And for what it's worth, I still feel that your present setup is undercharged based on the large delta T across the primary side of the evaporator.

Lastly, it is still quite impressive what you and your team have accomplished thus far.

[ultralo1]

Methodology! Thanks Michael, Thats the word i was looking for!

[Kevin Hotton]

Hi Michael,

Very interesting on the comparison of the flat plate efficiency vs. the tube-in-tube HX, and something that I've begun to suspect recently. Perhaps Polycold will soon be in trouble, because they have abandoned the tube-in-tube HX in favor of the flat plate on there newest XC Chiller unit.

It’s not that flat-plate HX’s are inherently inferior to tube-in-tube. Any HX (if it is to work well) must be designed appropriately for the application. To transfer heat in a HX requires: 1) area, 2) temperature difference. The key is how efficient is the area of the HX in transferring heat. The cascade-condensers in autocascade cycles are really partial condensers on the high-pressure side and partial evaporators on the low-pressure side. So both sides of the heat-exchangers have a high vapor fraction. Vapor (gas) is much inferior to liquid in transferring heat (that is the reason for the tightly spaced finned on air-cooled condenser – to greatly increase the heat-transfer area on the air side. If flow velocity is increased, heat-transfer improves (but with added pressure-drop). To increase the velocity through a flat-plate HX means reducing the number of plates, but doing that also reduces the heat-transfer area.

So the problem I am having is because I could not spec-out custom flat-plate HX’s. This might not be the case for Polycold. If I could I would want a flat-plate HX which was narrow and long in shape so that the cross-sectional flow area was reduced (increasing flow velocity) but long in length to still give the needed heat-transfer area. Compare these two flat-plate models from dudadiesel. The long 30-plate model is the only one offered that comes close to what I am talking about.

Kevin

[n00b 0f l337]

Oppositely to that though, your simulation constantly goes for a solid liquid flow at each captube. This is not necessarily required or always the case with a mixed refrigerant cycle.
When it comes to surface area, the differences in liquid to gas phase can effect the heat exchange, or trap liquid, or all sorts of other possible errors that using a coil HX simply eliminates.

[mytekcontrols]

Oppositely to that though, your simulation constantly goes for a solid liquid flow at each captube. This is not necessarily required or always the case with a mixed refrigerant cycle.

Very good point Adam, and I found this to be absolutely true when applied to downstream cap tubes (phase sep 2, 3, and on). In fact the liquid quality becomes so poor, that often times the final cap tube in the system is the same ID and length as the #1.

Edit: Because the first stage of an AutoC needs to move the greatest amount of heat, the #1 cap tube also needs to have the greatest amount of flow (large ID, shorter length). However using this same cap tube spec for the final (evap) cap tube will not yield the same amount of refrigerant flow due to very poor liquid quality at this colder stage.

Also it has been found that the down stream phase separators can be progressively less efficient, and in some cases be nothing more then a tee which allows a small portion of the available liquid to spill over.

[mytekcontrols]

Speaking of heat transfer in flat plates versus tube-in-tube HX's. I think the biggest problems are related to insufficient allowance for temperature difference across the HX due to the short distances involved from the inlet to the outlet side, and the inherent liquid hold-up problems, both of which tend to be a flat plate attribute when used in phase change applications. Kevin you appear to be on the right track when you suggested going with both a longer as well as less plate design, which would most likely improve on both of these aspects.

[Kevin Hotton]

Just a brief update. The autocascade is currently being used in another lab scale system. The refrigeration system had been very reliable in cooling and condensing a mixture of hydrocarbons to below -60C.

Kevin