Our Second AutoCascade

[Kevin Hotton]

I will be reconfiguring our first AutoCascade (a two phase-separator with no auxilary condenser) into a single phase-separator with an auxilary condenser. The first design was a great learning experience and it will be very useful data to help me select proper captube lengths and refine the exact "design-point" cycle I will try and match. I now have a general idea of the heat-transfer performance of the HX's I have to use (and have incorporated this knowledge into my simulation model). Based on the test results with our first AutoCascade, I am confident the cycle performance measured will be quite close to my simulation model.

Refrigerant mass fractions are 58% n-butane 38% ethylene (at compressor inlet). Note in the "Compostion" table that only about 5% n-butane (mole fraction) is present in the evaporator.

I am going to tweak this cycle for CPU cooling duty (target -80C, 300W duty) and will post cycle details for "over-clockers" that my be interested.

Kevin, Matt, and Joe

[mytekcontrols]

I look forward to seeing how this new system performs

On a slightly different topic, but one that might play a role in what you are doing, I think I was wrong about my earlier thoughts on HX orientation. Here is a quote from a paper put out by the University of Illinois titled: Investigation of Adiabatic Refrigerant Pressure Drop and Flow Visualization in Flat Plat Evaporators (by E. W. Jassim, T. A. Newell, and J. C. Chato -- July 2001)...

The optimal plate heat exchanger inclination for evaporation and condensation have been investigated by
Kedzierski (1997). He found that the optimal inclination for evaporative heat transfer in flat plates was the vertical
upward flow configuration. Gravity induces stratified flow in the horizontal flow configuration, which decreases the
heat transfer. If the evaporator was horizontal, it was found to have 60-75% of the vertical position heat transfer. For
condensation, however, it was found that the horizontal position was the optimal position for heat transfer. The
condensate film thickness becomes thinner in the horizontal flow configuration, which increases the heat transfer.
The horizontal position provided heat transfer 17-30% higher than the vertical downward flow configuration.

So according to what these guys discovered. apparently your use of horizontal orientation for the condensing HX's (Cascades) appears to be better for heat transfer. Whereas a vertical orientation for the evaporator is preferred, and exactly as you had it on your first AutoC unit.

[n00b 0f l337]

In using this program I'm finding a few things that I cannot grab a tutorial for...
Is there a means by which you've calculated pressure drops across the heat exchangers? Or have you set this to 0.

Similarly, with energy streams, is there a method for determining compressor/condenser numbers? Thank you.

[Kevin Hotton]

Hi Michael,

Thanks for the information and your enthusiasm. Actually the main reason I chose to install the cascade HX's nearly horizontal is that of refrigerant quality. On the low-pressure side quality changes from a "wet vapor" to a "less wet vapor" (partial evaporation -- so flow direction is very slightly uphill). I was afraid that if flow velocities were not high enough with the large cross-sectional flow area of the 20-plate HX's I was using (unlike the case with tube-in-tube HX's where their orientation is less important) that I could pool refrigerant at the bottom of the HX if they had been oriented vertical.

Kevin

[Kevin Hotton]

I input estimates for the HX's pressure drops. Usually 10kPa for a plate HX that is oversized like the ones I have been using. Probably 20kPa for more typically sized plate HX's and then anywhere from 10-to40 kPa for the tube-in-tube designs depending on heat load. I could refine these estimates if I delved into the fluid dynamic / heat-transfer equations, but I am too lazy to put forth such a tedious effort (until I have a fairly refined cycle design).

For the compressor you can enter a isentropic efficiency (the default is 75%) but I lower this to 70% for rotary or scrolls (or to performance data if available).

Condenser heat load should be calculated automatically (but you need to remember to specify zero-quality (saturated liquid) or better yet, see what the saturated temperature is and then add some sub-cooling to the condenser outlet state-point.



Kevin

[n00b 0f l337]

Interesting.
Coming up with some of these values is a quick headache
...

And just found another hole tab of values. Time to dig up that capillary tube chart. I'm assuming this "valve" is set to 100% open and then input the mass flow?

[Kevin Hotton]

No you don't need to specify pressure drop at the valve (they are modeled as constant ethalpy pressure-drops which are not exactly what occurs with captubes, but the error introduced is small) just specify compressor discharge pressure, evaporator suction pressure, and the pressure drop across all your HX's, and once you close the cycle loop, the program will solve for the pressure drop across the expansion valves.

Kevin

[n00b 0f l337]

Thanks Kevin, this is quite some piece of software.

I seem to have to specify a temperature to get any numbers, which seems a bit backwards. And no matter how I tinker I seem to see incredibly crazy discharge temperatures from the compressor. I'm assuming I'm doing something wrong. Would it be possible to discuss some of your normal steps in using the program?

EDIT: Interesting, don't seem to have to actually fill in the compressor duty or condenser specs, this throws it off. Just evaporator load and temperature. That seems good for finding out the numbers you want for a goal. Time to figure out mixtures.

EDIT: Whoaaaaaaa.
Made a single stage. Mind blown. Comparing it to raw data I have from a miny single stage. Verrrrry close. Predicted power consumption of compressor within 8%, temperature within 3C.

EDIT AGAIN: Trying to figure out how to get the "Set" function to work to link up a Cooler to a Heater, but never seems to work out. Or I get negative mass flows, etc. Any idea?

[Kevin Hotton]

nOOb,

the "SET" function is quite easy. If you are trying to match the heat load between a Heater and Cooler, you make one, say the Heater the "target variable" in the "object" list and "heat flow" in the "variable" list. What is important is where you started to build your cycle. I usually start at the compressor inlet (lable it 1) you enter inlet pressure and mass flow rate, and refrigerant composition there. The "set" function can only tranfer the value of a parameter that it already has a value for. For example in my autocascade cycle I had to define my cycle past Q-Csc for the program to calculate the heat-load that could then be assigned to Q-Hsh on the suction returning flow.

Hope this discription helps,
Kevin

[n00b 0f l337]

Thanks, not enough time to play with it I guess. I got SET working, but still having trouble coming up with the numbers to assign some requirements.
Interesting though. So for this project, do you have an evaporator in mind?

I've been doing some maths as far as evaporator function and keep coming back to that almost every evaporator I have seen used in the last few years (with one or two MINOR exceptions) was massively undersized from a volume point of view, and we've been basically scraping by on cooling refrigerant to a liquid phase and using that as a working fluid instead of properly.

[Kevin Hotton]

I have been comparing the two different AutoCascade designs we are trying. I have modeled them for the full scale 167kW cooling duty with a -75C evaporator. If I use the same design constraints (such as LMTD log mean temperature differences of 10C or greater) a compressor efficiency of 72% The two phase-separator (2PS) design without the auxiliary condenser does appear slightly better. Earlier I thought the one phase-separator (1PS) with an auxiliary condenser might be better, but it no longer appears so.

They are close in performance, but the differences I have found are:

1) total heat transferred over all (four) HX’s is 11% greater in the 1PS than in the 2PS design.
2) Compressor power is 5.7% greater in the 1PS design.
3) The average LMTD across all the HX’s is 12% greater in the 2PS design.

All these slight differences result in a coefficient of performance advantage to the 2PS design of 4.1%. The 2PS design will also use smaller HX’s. However, the 1PS design should be easier to tune (with only two expansion valves) and should also cool down quicker.

A normal 2-stage cascade with an evaporator temperature of -68C even with an increased duty of 175kW is significantly more efficient (over 15%).

Kevin

[ultralo1]

It will be interesting to see how the TXVs work on the system. I have found in the past when trying to run evaps in parallel, while using TXVs, it has been a challenge to get the system tuned. As one was hunting it would cause the other to sympathetically hunt also. You would have to make a change and let them settle down and stabilize. This could take up to an hour. To me the TXV alter too many variables at once for multiple evaps, but my experience with them in parallel is very limited.

[Kevin Hotton]

This is a preliminary design for an autocascade for CPU cooling. The duty is 300W and the evaporator temperature is -80C. The compressor was modeled as a rotary and needs to be about 20 cc/rev (a R22 unit from a 14000 Btu/hr window AC). If the more common 12000 Btu/hr (17 cc/rev) is used, duty would be reduced to 255W. Refrigerants used are n-butane and ethylene.

Kevin

[Kevin Hotton]

I revisted the above -80C design with the goal to make it more practical. A loaded -80C evaporator using ethylene (R1150) requires suction pressure of under 3-bar. This leads to a large capacity (but lightly loaded) compressor and higher pressure ratios reduce compressor efficiency. By relaxing the design to a -70C loaded evaporator (also 300W) the design becomes more favorable in several ways:

1) Pressure ratio (PR) reduced from 5.0 to 4.5 (the compressor will operate more efficiently as standard AC cycles use PR around 3)
2) Suction pressure increased from 2.6 bar to 4.0 bar (closer to standard AC suction pressure).

The volumetric flow at the compressor inlet is 1.87 m3/hr. Using a volumetric efficiency of 0.87 yields a 10.2 cc/rev sized compressor. So something slightly larger than the 9cc/rev model Michael used (in his "morph" the AC unit thread) would be needed for a full 300W duty.

Below is the Aspen Hysys model of the design. The refrigerant charge is 50% n-butane 50% ethylene by mass.

Kevin

[n00b 0f l337]

I see many of our questions in system sizing often arise backwards.
For years Gray Mole and others have stated the "Evaporator backwards" theory. Which is absolutely correct.

But much of the time we work with what we have on hand or can source from ebay or otherwise. Based on this, we get a 12,000 btuh compressor and try to maximize with it.

So I have to ask the question, obviously rigging it as I go, "What's the best you can manage with a 12,000 btuh compressor, assuming r410a compressor; to hold 400W".

Just trying to keep the discussion ball rollin.

[Kevin Hotton]

Based on the Rechi compressor website, it appears that 410A models of 12000 Btu/hr are about 12cc/rev (very convenient ratio of 1cc/rev per 1000 Btu/hr).

So above 10.2cc/rev for 300W, and 12/10.2*300 = 353W. Note, the suction and discharge pressure do not require a R410a compressor. Using Michael's model (6500 Btu/hr R22?, with its 9cc/rev compressor) would suggest a 12000 Btu/hr unit would have a 12000/6500*9 = 16.6cc/rev (seems like I have seen mentioned on this site 17cc/rev often). So 16.6/10.2*300 = 488W.

Wow, so a larger unit (not 410a) 12000 Btu/hr AC unit is big enough for over +450W duty at -70C.

Kevin

[n00b 0f l337]

Interesting. It does seem like these R410a systems have much lower displacements.
Not exactly the direction we've all wanted to go.
But they do let you run the more extreme pressure ratings.

[Kevin Hotton]

These are a couple photos of the HX's used in the autocascade. The three HX's (top to bottom) are cascade condenser, auxilary condenser, evaporator. The HX's will be installed in a modified window AC unit (10,000 Btu/hr). This will be cooling a methanol flow to -70C with a duty of 300W. If this was to be used for CPU cooling the evaporator coil would be eliminated. The autocascade cycle being used is the -70C cycle I posted earlier.

The auxilary condenser (12-ft long) and cascade condenser (16-ft long) are two 3/16-inch tubes within a 1/2-inch tube. The evaporator (14-ft long) is 5/16-inch tube within a 1/2-inch tube. I used drawn 1/2-inch tubing (but annealed it with a rosebud torch) before bending the coil.

Kevin

[n00b 0f l337]

Very interesting use of the hard drawn pipe. Very smart, but seems like quite a bit of work.
What kind of pipe did you use? On the sides it looks as if they'res an imprint at the bend...

[Kevin Hotton]

I used a Swagelok tube bender. The marks are where the hold-down clamp grips the tube. I bought 20-ft straight lengths of drawn copper ACR tube. The 3/16 and 5/16 inch OD tube was in a 50-ft coil.

Kevin

[Kevin Hotton]

We ran the autocascade for the first time today. No insulation or load, but it did get down to -80C at the evaporator inlet. The design point is -72C with a 300W load, so I think I am close. The HX stack is (top to bottom) cascade-condenser, auxilary condenser, evaporator. The charge I used was 3-oz (wt.) of n-butane and 3-oz (wt.) of ethylene. The compressor discharge pressure was 320-psig and the suction was 41-psig. The compressor was running a bit hot at +120C but it never shut off on thermals. I plan on blocking off the air vents on the unit's cover -- leaving open only the one on the side by the compressor. This will cause the condenser fan air to be pulled around the compressor and should help with cooling. Still the compressor temperature is not that above my simulation predition of 114C. After charging static pressure was about 250-psig (after a few start/stop runs the static pressure has settled at 225-psig). I have a small expansion tank with 10-ft of 0.031-inch captube connecting it to compressor suction and this limits the rise in discharge pressure during start-up. At warm start-up discharge pressure stayed below 400-psig and than slowly settled at 320-psig. I am expecting with insulation, that the discharge pressure will be closer to the design-point value (250-psig). Also, remember this is a R410a compressor so 400-psig is not unusual. For those interested, the captube on the phase separator is 8-ft of 0.040-inch ID and the evaporator captube is 8-ft of 0.031-inch ID.

Kevin

[n00b 0f l337]

Congrats, although I see the aux setup next to the evaporator as a slight design flaw...

[Kevin Hotton]

Today I added insulation. Actually, it is just a "box" made up of 1-inch thick neoprene glued up with rubber cement. I intended to fill the interior with perlite (or some form of expanding foam) but have not. As a result the HX coils create a refrigerated space around the phase-separator. When I run the unit now, the discharge pressure settles at a much lower pressure (only 200-psig vs 320-psig), but without load the evaporator is actually slightly warmer (-75C). I believe what is happening is the phase-separator is being sub-cooled to below the design point temperature and this is creating more liquid/less vapor in the separator. The low discharge pressure (along with the unit's cover installed and selected vents blocked) resulted in a very comfortable 65C compressor temperature.

update: I did add two-part spray foam insulation to fill the interior of the heat-exchanger coils to isolate the phase-separator from the chilling effect and this has worked well

Kevin

[mytekcontrols]

When I run the unit now, the discharge pressure settles at a much lower pressure (only 200-psig vs 320-psig), but without load the evaporator is actually slightly warmer (-75C). I believe what is happening is the phase-separator is being sub-cooled to below the design point temperature and this is creating more liquid/less vapor in the separator.

I agree, this seems like a very plausible explanation. You will probably need to add more ethylene.

[n00b 0f l337]

Additionally the lower discharge pressure means less saturation is occuring, as well as a higher vapor pressure coefficient. And less mass flow.