I know many people are looking at He recovery systems. I recently analyzed
my data from our first full year of operation and figured I'd share the details
here for those interested. I've also included some details about our setup and
things I've learned that would have been nice to know in the planning stages.
I hope this is helpful.
Quick Summary
5 superconducting magnets + a CryoMech recovery system
Delivered 870 L over 1 yr in 23 magnet fills
Recovery = 97%
Annual costs:
Annual He loss = ~30 L/yr = $700 (+ 10% - ~$800)
Expected mainten. divided out by year = $2.4k (+ 10% - ~$2.6k)
Annual electrical usage = $3.6k (+ 10% - ~$4k)
(+10% to account for small errors, price increases, etc.)
$3.89 per L delivered (ignoring electricity)
$8.79 per L delivered (including electricity)
We were paying $23 per L delivered (2021, Airgas)
That's a savings of $21k per year.
Total capital costs to purchase/install the system - $309k
That's 15 years to recoup the capital costs (sooner, as He
prices continue to rise ...)
Note: We have one more magnet that will be added to the system later
this year. It needs 40 L every week(!), so the system will be running at
almost 100% capacity. This results in more frequent maintenance, but
drops the cost per L to $2.39 ($5.03/L including electricity), reducing
the time to recoup capital expenses to 11 years.
Details about our system:
Note: This gets rather long winded at this point - you've been warned.
CryoMech recovery system consisting of -atmospheric pressure recovery bag
My bag holds roughly the volume of 1 compressed gas cylinder (i.e. the
equiv. of 8-9 L of liq He)
One advantage of a bag vs. a rigid collection tank is no observable P
fluctuations at the magnets when emptying the bag or collecting the
plume from another fill.
-8 medium P storage tanks
This holds roughly the equivalent of 125 L of liq He at max. pressure
(400 psig).
-recovery compressor
This moves the He from the bag to the storage tanks. It's turned
on/off automatically based on the size of the collection bag.
This is much louder than the pumps and cold heads in the purifier and
LHeP. I stuck it in a side room so I can shut the door and reduce the
noise in the NMR lab.
-automated purifier
This removes any N2 and O2 that might be in the collected gas. It
provides UHP grade He to the liq He plant. This is the most
temperamental part of the system ... it works great, but has to
warm up for purging the trap more often than I expected and that takes a
lot of time (several hours to warm up and purge and then ~8 hours to
cool down and get back into steady state purification mode). Also, one
of the vacuum spaces around the cold head has needed to be pumped down
about every 6 months.
-LHeP22 with 150 L collection dewar
The nominal/advertised liquefaction rate is 22 L/day, but I've
been getting 25 L/day for the past year. The only problem I've
had is the interface periodically locks up and needs to be power cycled
(appears to be due to the Windows + LabView interface).
Details about expected maintenance schedules and costs, etc. are below.
Misc Details
The He collection circuit is built from 1” capped
and cleaned Cu tube connected with Pro-Press fittings. Pro-Press has a
specific o-ring that's rated for He; these had to be ordered
separately and swapped out for all the standard o-rings during assembly.
Our physical facility people prefer to use Pro-press because it's
faster and cleaner than welding (e.g., I had our phys. fac. people weld
some brass KF25 nipples onto Cu pipe and the amount of black crud I had
to clean out of the pipe was astounding!) We have two magnets on the 2nd
floor and 4 magnets in the basement (i.e., 3 magnets in the same room as
the recovery system and 3 magnets further away), so this required a lot
of Cu tubing.
I was initially skeptical about using Pro-press fittings for
this given how easily He leaks. Once the main circuit was built, I
pumped it down to ~25 mTorr and then monitored the P over ~2 weeks and
saw no decrease in P. I've also looked for evidence of leaks in
our fill data (see below) and everything looks great this past year.
I copied Yale's design for the manifolds connecting the
magnets to the collection circuit (KF25 fittings + flex lines, back-pressure
regulators, Omega 7611 rotameters, etc.) The Omega 7611 rotameters work
really well! They make it really easy to detect when the magnet is full
while capturing the plume. Through trial and error, I've found a flow of
~4.0-4.5 on the rotameter is just a bit slower than the rate my recovery
compressor empties the collection bag (this is particularly helpful when
filling magnets in other rooms).
*** Plan for a good floor scale to monitor the filling of
the xfer dewar (and to periodically monitor their boil-off). Somehow
this completely escaped me during my planning (I guess I was imagining
that I would simply roll the dewar to and from the stockroom and use
their floor scale ... such a terrible idea!) If you can, make sure
you have room for a scale with a large shallow ramp; due to my poor
planning, I could only fit in a small scale with a short/steep ramp and
it takes some wrestling to get the xfer dewar on/off the scale. Also,
get a scale with high resolution ... my scale only reads +/-0.5 lb
which ~ +/- 12 L. It would be much easier to monitor the filling of xfer
dewars with better resolution.
Practical Data: (averaged over Mar '21 - Mar '22)
What is the transfer loss when filling a magnet?
o It ranges from ~3/4 L - 2 L per magnet fill.
o I was frustrated that no one had a concrete answer for this,
so I used a full xfer dewar and cooled down a warm xfer line until I got
a nice flame and then vented the dewar. I repeated this with several
warm xfer lines until I saw a change of 1 lb on my scale. This gave
~0.75 L of liq He for cooling a warm xfer line and venting the pressure
in the dewar.
o After 1 year of fills in the NMR facility, I observe an
average transfer loss of ~0.8 L per fill.
o However, the data from the other magnets give transfer
losses ranging from 1-2 L per fill. This is partly due to the high back
pressure for those magnets further from the recovery system (you can see
the He escaping from the inlet during the fills!), but also due to
filling technique – my colleague who fills the FTICR magnet has
a rather more cavalier technique than I do and sprays large amounts of
liquid He out of the xfer line while fiddling with the inlet port ...
I'm pretty sure that 2 L xfer loss could be dropped to ~1.0-1.2 L per
fill with a more careful approach.
How much liquid He needs to be condensed for every L that goes in the magnet?
(i.e., what is the vaporization loss or "flash" during a transfer?)
o For magnet fills, I see 10%-20%. This depends strongly on
how fast the fill is made. I fill my magnets at a reasonable rate and
get ~20% flash while my colleague fills his FTICR magnet at a much
faster flow rate, but consistently only has ~10% flash during the fill.
When averaged over all 5 magnets across 1 year I get a flash volume of
~18%.
o When filling the xfer dewar, I see 25%-35% flash. Averaged
over 1 year it's ~30%. I suspect this higher number is due to
the pressure difference between the collection dewar and the xfer dewar.
I've noticed that I see much more flash if the collection dewar
is at 6 or 8 psig vs. 3 or 4 psig. However, if the pressure is too low
(e.g., 1-2 psig), the transfer rate is quite slow and that also leads to
a very high flash %. I could solve this by using a tank of UHP He to
push the He out of the collection dewar like I do when filling magnets,
but with the cost of UHP He, I just use the pressure from the purifier
to fill the xfer dewar and live with a bit higher flash.
o You need to take both of these flash %'s into account when estimating
the volume you will condense over the course of a year. For example,
13 weeks of boil off in the NMR facility - ~70 L
Transfer losses for 3 fills bump this up to ~73 L
Due to ~20% flash when filling the magnets = I'll push at least 91 L out
of the xfer dewar (say 95 L to be safe).
Due to ~30% flash when filling the xfer dewar = I’ll push at least
136 L out of the collection dewar.
i.e., to put 70 L in the magnets, I need to condense 136 L of liquid He.
Note: only ~3 L of that transfer loss is truly lost ... I collect
the plume while filling the xfer dewar and the magnets, so all of the
flash just goes back into the system. But, the you need to keep this
extra volume in mind when estimating how long it will take to condense
the liquid for a fill and when estimating the maintenance schedule or
how much electricity is used, etc. ***This also means you need to have
more He in your system than just the liquid that you want to end up in
the magnet!
How much He is lost when filling a warm xfer dewar?
o This depends a lot on the technique. I tried the whole
pre-cooling with liq N2, purging it, pumping, etc. and I have no desire
to jump through those hoops again. When we got our second xfer dewar, I
figured I'd try cooling it with just liq He ... as long as I'm collecting
the plume from the fill, I figured it didn't matter how much boiled off.
o I used a very slow flow for the first 3 hours (i.e., I
pushed in ~10 L over 3 hours).
o Once I could see the weight of the xfer dewar increasing, I
increased the flow to the normal rate.
o By the end, I had collected 58 L of liquid in the xfer dewar
and the equivalent of ~22 L of liquid in the recovery system (this was
from a commercial dewar that started with 80 L of liquid). That's
only ~28% lost to flash ... I'm honestly shocked at how
well this worked; I was expecting to lose 50% or more.
Cost Estimates
With good values for transfer losses, flash %, average
boil-off rates, and your annual fill schedule you can estimate the total
V of liq He that must be condensed. ***You also need to periodically
top off the xfer dewar to keep it cold as well as run the LHeP
periodically to keep the collection dewar cold. This adds more hours to
the system, affecting the estimated mainten. schedule and electrical
costs. Note: the CryoMech LHeP has an on/off mode that runs at a 40%
duty cycle when not actively condensing liquid He, but the purifier
needs to be turned on/off (with the associated long cool down time).
Total boil-off (5 magnets) = 12.5 L/week = ~700 L/yr
Transfer losses = 23 fills x 1.2 L/fill = ~30 L/yr = ~730 L/yr
20% flash when filling magnets = ~870 L/yr delivered
30 % flash when filling xfer dewar = ~1250 L/yr condensed
Plus ~675 L to keep the xfer dewar cold between fills = ~2000 L/yr
At 22 L/day (+ 40% duty cycle when not actively condensing) - running
the LHeP ~4800 hrs/yr and the purifier ~2700 hrs/yr.
Estimated average annual maintenance cost:
Automatic purifier
Replace oil adsorber ($2k) every 25,000 hrs = every 9 yrs
Replace cold head ($6k) every 40,000 hrs = every 15 yrs
LHeP22
Replace oil adsorber ($2.5k) every 25,000 hrs = every 5 yrs
Replace cold head ($6.8k) every 40,000 hrs = every 8 yrs
Misc costs (ground pin, purity sensor, etc) every few years.
Average yearly cost - ~$2.5k/yr
Note: if the oil adsorbers are replaced on schedule, the cold heads can
last much longer than 40,000 hrs. Monitor the liquefaction rate to
determine when performance declines.
We don't actually pay our electrical bill, but some people
need to, so I include an electrical estimate here for comparison. My
university has a contracted rate of $0.07/kWhr. The automated purifier
is 3.1 kW and the LHeP22 is 10 kW. Using the estimated run times from
above - ~$4k/yr in electricity for my facility.
To put all this into perspective, I calculate the cost per L
"delivered" (i.e., what I push out of the xfer dewar) as
that can be most easily compared to the cost of a commercial dewar. For
example, a "100 L" liquid cylinder from Airgas never
arrived with more than 85 L in it, so the cost of the dewar + shipping +
demurrage gives me a rate of $23 per L delivered from Airgas.
Interestingly, the cost per L (liq. equiv.) from a compressed gas
cylinder of regular purity He from Airgas is ~$21, so there's
not a big savings using compressed gas over liquid He (UHP, of course,
is more expensive at ~$31 per L of liq. equiv.) Note: these are all
2020/early 2021 prices; it’s probably more expensive now.
The numbers above give $3.89 per L delivered ($8.79 if
including electricity) vs the $23 per L delivered that we were getting
from Airgas. That’s an annual savings of $21k (based on the
number of 60 L and 100 L we were purchasing – some of which had
significant amounts of liquid left after a fill!)
Total capital costs = $309k
CryoMech system = $220k
Renovation costs = $57k
Misc = $32k (manifolds, plumbing, 2 liq He dewars, etc.)
With a savings of $21k/yr we recoup the capital costs in ~15 years.
Using the estimates above – I deliver 870 L/yr and lose 30 L/yr - 97% recovery.
Looking at the data from the past year, I observe a recovery of 95%.
However, some of my early fills weren’t very optimized. The past
5-6 months give recovery ~98%.
Misc Info
I was concerned about leaks in the system, so I look
carefully at my numbers for that. Specifically, I looked at the total
volume of He in the recovery system (collection dewar, xfer dewars,
storage tanks, with an empty collection bag) just before and just after
every magnet fill and every xfer dewar fill.
Using the average boil off rates to predict the volume that
should have gone into each magnet vs. how much He was removed from the
recovery system gave me the transfer loss for each fill. Averaged over
23 fills for the past year = 1.2 L per fill.
Similarly, using the average boil-off numbers, I can predict
how much He I should be collecting between fills and compare that the
total He in the recovery system just after a fill vs. before the next
fill. Totaling this difference (rather than averaging it) over the
course of the year should reveal any significant leaks as an overall
negative change. However, I collected 2.6 L (liq.) more than predicted.
This could be due to the He that's injected into the system from
the push gas with each fill, but could also be related to the large
fluctuations in boil-off as well as my large errors in xfer dewar V due
to my scale resolution. I still think this is good evidence that I
don't have a significant leak in my system.
For reference, my average boil-off rates and fill schedules are
NMR facility (1 AlOx + 1 Magnex + 1 Bruker) = 5.4 L/week (4 fills/yr, every 13 weeks)
SSNMR (Magnex) = 3.0 L/week (3 fills/yr, every 17 weeks)
FTICR (Varian) = 4.0 L/week (8 fills/yr, every 6.5 weeks)
PPMS (Quantum Design) = 4.0 L/day = 28 L/week (not on
the recovery system yet ... expect to fill this each week or every
two weeks)
Xfer dewar = 1.3 L/day = 9.1 L/week (topped off after
each fill, occasionally topped off when there are long gaps between
magnet fills)
Xfer dewar with heater = 2.1 L/day = 15 L/week (will be
used to fill the PPMS, currently keeping it cold by topping it off once
a month)
I hope this is useful for someone.
-Scott
---
Dr. Scott Burt
Teaching Professor
NMR Facility Manager
Department of Chemistry and Biochemistry
Brigham Young University
C414 BNSN / C008A BNSN
Provo, Utah 84602-5700
Phone: (801) 422-2404
chembio.byu.edu/nmr-facility <https://chembio.byu.edu/nmr-facility>
email: srburt_at_chem.byu.edu <mailto:srburt_at_chem.byu.edu>
Received on Tue Mar 15 2022 - 05:03:45 MST