Using a Centrifuge to clean and dewater my WVO
I have been using a Dieselcraft centrifuge (designed originally for lube-oil bypass filtering, price= $200) to filter my waste vegetable oil (WVO) and it works great. I no longer need to buy replacement filters, and it dewaters the WVO (which no filters can do - including "water block" types.). I put the WVO sucked right from the dumpster through a 100 mesh screen into the rig, no settling needed.

There is a forum with much more information about this, and you can ask questions at:

The WVO needs to be heated to >120F and then pumped through it at 90psi and 0.9gpm. It spins at around 7000 rpms. The way I have the water heating elements, it is removing water both by centrifuge action (the best) and by Flash evaporation. I have sent samples to a testing lab and it showed good results, both for water and particle removal. Its easy to clean the rotor of the goop it removes. One operating trick I have learned is to capture the output of the rotor from the instant I shut off the pump, since it starts to dump WVO with water and dirt as it slows. I run it all in a single barrel, and from my testing 4 passes seems to get most of the particles out, so it takes 2-3 hours for a 40-50 gallon batch. The longer you run, the more it removes. This works as a mobile rig very easily and is what I intend to use it for, as well as at home.

With any version of this you should make provisions to capture the output of the centrifuge right when you turn it off. The output at that time is often very dirty besides containing up to 4 oz. of water (if there was any water.) By doing this you are removing at least part of the water as a true centrifuge does, which is better than the Flash evaporating that this also does. This can also make it so you can avoid the mistwash step.


I used a closed top barrel, put upside down so I could use the 3/4 threaded fitting, then cut off the top (which used to be the bottom.) Note that I tilt the barrel towards the output so that any water on the bottom gets sent into the centrifuge. If you don't suck from the lowest point, there could be a pocket of water that settles out and keeps re-contaminating the dewatered WVO. A cone bottom tank would be the best, but this was a lot cheaper. Here is a good way to test if you have a good barrel drain and tilt, pour an inch of water in the bottom and see if it all runs out the drain pipe. The suction should come from the lowest point, and the pipe should go continuously lower until the pump. Otherwise you could get a pocket of water in your pipes. This the most common problem I see now that many have tried to copy my design and "improve" it.

(edit:)Now I have a Sandy Brae water in oil tester, and here are the results from a centrifuge run I did recently. Average quality WVO source, FFA=2.5%, settled for 4 months. Began at 700 PPM and end 270 PPM with sandy brae, also measured every 30 minutes with my
Homebrew capacitive water in oil meter to make this chart, and temp. compensated.

Reasons that I moved the centrifuge below the rim of the barrel (see 10-14-06 pics below in the timeline), was so it could be liquid tight if I sealed where the feed line perforates the drum with a bulkhead fitting, and using a solid lid with gasket, it could be run or at least carry a load of WVO while travelling down the road. When I run a batch that's less than 45 gallons, like around 20-30g, it can splash a little and I set a lid on top, with it partially open by the centrifuge for water vapor to escape. A piece of window screen also works good to prevent splashes or debris getting in there. The only debris I ever got was insects flying into the WVO. The centrifuge placement I used is also a better height for the window screen trick, which still allows the water vapor to escape.

Many have asked me why don't you use 2 barrels so you are sure that all the WVO goes through the centrifuge each pass. I tried it, and so have many others, it doesn't reduce the number of passes, and it makes it much harder to run since you need to be there each time before the barrel runs out to prevent pump damage and capture the output of water and goo from the centrifuge. And you need to clean the rotor much more, each time it runs out. Its not worth it. Thats too much babysitting for me, I like to start it up and come back 4 hours later when its done.

I kept all the text and dates so you can see its development as I went along:

25 September 2006

I just ran my first batch of 15 gallons of WVO straight from the restaurant barrel, 100 mesh (149 micron) screen on sucker, through the centrifuge with no other filtering and it works great. Easy to clean, no more filters to buy. Runs at 90psi, 1 gpm.

Here is my rig, it goes pump>heater>centrifuge>barrel:
This was before I added the fittings on the bottom of the barrel for the pump intake.
The pump is a free power steering pump off a 84 Mitsubishi montero, hooked to an old 1725rpm 3/4 hp motor I had. As small as 1/4 hp should work, this is what I had laying around (I got a better pump later, pics later in this post):

This pump was easy since it had a rubber line going to the reservoir which I just turned upward and attached my inlet hose:

This goes into a 4500w 220v water heater element run at 110v (=1125w) in a tee in the bottom of the 1.25" vertical pipe. They are longer than an similar wattage 110v element, which means lower watt density, which means less chance of thermal polymerization (burning) of the WVO. The oil started at 50 F and got up to 110 F first pass. An inline heater is more efficient since you don't have to heat up the entire barrel, it adds 20-30F right before the centrifuge. If you had to heat the entire barrel that much more means it takes much longer, and you get more heat loss. Plus you get better heat transfer with flow across the element, and its less likely to burn the WVO on the element.
I did 3 passes with decreasing amounts filtered each pass, lots in the first pass, with almost no black sludge 3rd pass.
Here is the black gunk that stays in the rotor:

You guys working on the "holy grail" of mobile flash evaporator and filtering might want to look into this since with a little higher temp it might do both. Or you could do a flash or vacuum evap. then right into this. At 120 F I was seeing small puffs of steam or atomized oil when pumping the hot oil through 2 small orifices at 90psi. This oil was water-free to begin with according to a pan test.

The oil I started with was black and you couldn't see through it at all. And here is the sweet iced tea drinkable looking WVO after only 15 mins for 15 gals:

Here is the makers website with a diagram of how it works:

Its very easy to clean, you open it, take a nut off the rotor, and wipe inside parts with a rag. It makes a mess on the rag but not much else.

I think you could clean lots before the rotor fills. The dirt is really compressed. I could have done hundreds of gallons of the oil I had without cleaning it. But I wanted to clean it often just to see the goop.

It needs specific flow rates to work. The dieselcraft website gives info, but its about 0.9 gpm.

Design and operating this rig:

For the pump, you want a small gear or vane pump. I almost bought an Oberdorfer 1000BR gear pump on ebay for $80 until I realized that power steering pumps are the right size pumps. Most pumps are too big, you want a small one, this one is about 3"x3"x3" not counting the pulley and reservoir. You also want it to be easy to mount and to seperate the reservoir to input directly to pump. Other models with the same pump I used is Dodge D50 and mitsubishi 2.5L truck from 80-89 and Monteros from 84-89. Test your pump with open flow, it should be about 1.5 gpm. This is the right open flow which will magically drop to about 0.9 gpm at 90psi. If its not the right flow, change the pulley size on the motor. My motor pulley is small so its reducing the slow 1725rpm motor some. Or you could use a variable speed motor if you wanted to be really cool.

Turn on the pump to fill the lines before turning on the heater! I turn it on for just a few seconds, it self primes quick, until the first oil drips from the centrifuge, then shut off. You don't want cold oil going through. Then I turn on the heater element, wait about 1 minute, then start the pump. When you hear it whine you know its working.

For the 1st hour, I use both 1125watt inline heaters. This makes the first pass WVO going through the centrifuge about 120F. After 1 hour, I usually stop, capture the output and clean the rotor. Then when I restart I use only 1 of the heaters, which heats the WVO slowly to 160F by the end when I am done after 2-3 more hours.

Future ideas to test:
This is so close to a flash evaporator already, that I designed it to have a second 4500w heater element at the top T.

Idea #1. put a 3 way valve on just before the centrifuge, run it through a seperate orifice several times into the same barrel until its hot and flash evaporated. Then flip the valve to filter.

Idea #2. About a pint of water may be trapped in the rotor along with the gunk. Try putting a quart jar under the output just when I shut the centrifuge off to see if any water comes out. (edit- see page 12, this idea works,catches about 4 oz., it may be possible to remove sugars/acids without a mistwash if you capture the output at pump shutoff)

Idea #3. Run this thing hot enough to do the flash evap and centrifuge in one pass.

I had 2 heater elements in it at first, when the top one sprung a leak and sprayed a fine column about 30' straight up producing a warm golden shrimp smelling oil shower. I took it off until I solve the gasket issue.

29 September 2006

EUREKA! (is that what a wizard is supposed to say when an experiment succeeds?) I ran the same 15 gallons of oil which now after mistwashing had lots of water, visible water, left in there on purpose to see whether it can remove it. Mist washed in the same barrel the centrifuge runs in.

I hooked up a 2x6 concentrating parabolic solar panel to add the equivalent heat of my 2nd 1125w heater element which I don't have installed until the seal is fixed. Plus I just like solar and had it laying around.

Right away it formed a terrible emulsion, with 1" of beer foam looking head, and I thought I just ruined the cleanest WVO I had seen (clean since it had previously been through the centrifuge for about 2 hours = 6 passes.) The milky emulsion cleared up in about 15 minutes of running. It started puffing out many clouds of steam really quick at about 120 F, much more than the first time I ran the oil. Note -it had not puffed any steam after 15 minutes into the original run a few days ago, which is another thing that made me suspect this was steam and not atomized oil. It also fits Tim's description of how his flash evap. steam releases, you can hear little random puffs just before each steam cloud is visible.

So I left it run for about 2 hours, it slowly got up to about 200 F, and I waited until it was not puffing anymore, then did the pan test. No bubbles at all, EUREKA! no water. This thing just might be the holy grail! More testing to come...

07 October 2006

I just ran my second batch through this centrifuge, and it worked great. This batch was 30 gallons, totally black and smelled like french fries, and I ran it for about 30 minutes. The amount of gunk in the rotor was large, around 1/4" thick all around, but still easy to wipe out, kind of a plastic mud texture to the gunk.

I added the pressure relief valve suggested earlier in this thread from graingers. It popped off and released some pressure at first with cold oil since at startup the pressure is higher, and then closed and didn't pop again.

11 October 2006

I solved the sealing of the water heater elements. Use a 1/16" thick by 1.25" ID metal washer first on the element, then a 1.5" ID washer with a 1/8" x 1.25" O-ring inside. The bigger washer acts like a retainer, preventing blowout. This seals great and has handled the high temp and pressure used with my rig.

So now I am using 2 - 4500 watt heater elements which gets the WVO hot real quick and makes both the centrifuge and the flash evaporator functions work better.

14 October 2006

I had an email asking for more detailed pics so here are some more. Centrifuge in the barrel. The Tee before it is the pressure relief valve with a plastic elbow over it to direct the flow down:

Here is the first heater element:

Second heater element (showing the metal washers that hold the o-ring in place):

Parabolic concentrating 2'x6' solar collector (not needed with the 2 water heaters, but fun!)

Pan test for water

Here is how I do my pan test for water

Smear a finger of wvo accross a fry pan (cast iron preferred)as a temp check.

Keep the sample of wvo to be tested handy. Enough for 1/4"-3/8" thickness covering the bottom works best.
Heat the pan on high temp until the smear begins to really smoke then pour in the sample.

Do not pour in a sample with any visible water. If water droplets are visible no testing is needed. There is water present in your sample. Visible droplets of water will spatter hot oil out of the pan and may cause burns or fire.

Look closely at the bottom of the pan where the oil meets it. Are there very small bubbles forming. This indicates some suspended water. The number of bubbles indicates how much water is present in suspended form.

Removing suspended water in 1 pass
23 October 2006

A major development to report today. This rig does remove suspended water as a true centrifuge! (but it keeps it in the rotor) With only a simple change to how you operate it: you must capture all the output right after turning off the pump. Several times half the rotor contents were dirty water! (even while the output was close to passing the pan test) It will remove and collect up to 4 oz. water each time you do this. This makes for very fast dewatering, and may solve the issue of requiring a mistwash beforehand which has been discussed so much.

I ran tests today which are getting sent to the lab for testing. I started with 12 gallons right from the dumpster, and on purpose sucked up some crud from the bottom of the barrel. And then I added 1/4 cup vinegar to make sure we got lots of acid, and then I sucked up a thick emulsion layer in an old settling barrel that wouldn't seperate after about a month of heat & settle. These are all things I would normally avoid but I was going for bad conditions. Then I ran it through all the rig without the centrifuge circulating it well and suspending it all in a nice mess. Saved a gallon of this for testing labeled #1. Saved 5 gallons of #1 to process later below. This was dangerous to do a pan test on it pops so bad. And the water/acid is suspended, it hasn't settled out after 12 hours of heat & settle.

Ran the centrifuge only (no mistwash) for 4 passes on 5 gallons of this mess. Labeled as sample #2. The rotor captured about 2 oz. water with acid PH. And removed lots of black goo. It removed all traces of water in the pan test.

Then cleaned out the rig, and pumped into it and heated the other 5 gallons I had saved of #1 to 120F, and did a mistwash with a mistpro nozzle, adding about 1 gallon of water. Then I ran the rig without the centrifuge for about 15 minutes, which turned this into and even worse emulsion looking like chocolate milk, and also really washed it. Then I let this settle only for about 30 minutes, and drained the free water from the bottom which looked like white milk, and had acid PH. There was still a large amount of water suspended and on the walls of the barrel. Then ran the rig for 6 passes on this mess. Lots of steam puffing out, and when I stopped midway 2 times, the rotor contained 4 oz. of free water which I captured the output when I turned it off. And about the same amount of black goo. Tested water free with the pan test, labeled samples as #3.

This is some very interesting results which leads to a new way to run this where you could remove the water much faster which I will test next time: Run 1 pass, empty the rotor of water, repeat until no more steam puffs which in every case has been a simple and accurate indicator that it will pass the pan test.

This also means with any version of this you should make provisions to capture the output of the rotor right when you turn it off. The output of the rotor at that time is often very dirty besides containing water (if there was suspended or free water.)

Operating like this could lead to the holy grail - an at home and mobile version that doesn't require a mistwash.

24 October 2006

I just ran the first test of my new capture method, on my bad #1 oil from above that I had saved, which had lots of suspended water according to a pan test. And the filtered WVO passed the pan test with no bubbles after 1 pass at only 120F temps. And the stuff I captured from the rotor is horrible on the pan test, very crackly and dangerous, and I see some small globules of free water in the bottom even.

Mistwash Rig
24 October 2006

Note after testing many batches with my TDS meter (below) I have shown that the centrifuge removes the contaminants so well that I don't mistwash any more. Here is what I built to mistwash some very bad WVO I got. I use the same centrifuge barrel shown above, and this drum lid $7.28
from grainger.

Then I drilled a hole in the top center and stuck this 4759T23 4gph mist head in it $22.14 from
here at mcmaster

Just add fittings to connect to a garden hose. I mistwash until the water comes out clear. I am testing using a TDS (total dissolved solids) meter on a mini-mistwash of about a cup of the WVO to determine if a mistwash is needed and if its removing anything. Most of the time I have determined that a mistwash isn't helping and is not needed.

I hook a clear braided hose to the valve at the bottom, which goes straight up the side of the barrel. After I have mistwashed for 30 minutes and there is an inch or 2 of water on the bottom, I open the valve and water goes into the hose and up the side. Then when the water gets high enough it comes out the hose automatically maintaining whatever level I choose by how high the hose is. I clamp the hose in different positions onto the rigid centrifuge pipe that goes up the side. You could use any pipe or board for the same function. This is nice since it allows adjusting for different batch sizes.

When its done, I close the valve, let it sit overnight for good separation, then lay the hose below the barrel on the ground, and use the same valve to drain all the water off the bottom by watching when WVO starts to come out.

TDS meter testing to decide whether to mistwash
17 November 2006

I just completed the first tests with
these meters, and they are very sensitive. I think the TDS meter is best since its showing acids and salts and soap. The TDS meter is sensitive enough to give a reading of 150ppm with only a few grains of salt added to a glass with 3/4" of distilled water. (distilled = 0ppm.) Totally repeatable, and it also responds accurately to a tiny amount of vinegar, and a single drop of soap (reads 150ppm) added. And it works just as reliably with my tap water which starts at 60ppm TDS, the same tiny additives increase it to around 200ppm. I think TDS is much better than the PH meter, which has trouble with repeatablity, and drifts around, and only detects the vinegar.

I then did mini mistwashes on several batches of my WVO, with samples that were straight from the dumpster, as well as mistwashed and centrifuged batches. They all were clean, 0-30 ppm TDS, which proves I have a good WVO source, and could probably avoid the mistwash.

Here is how I do a mini mistwash, I add 1/2 cup WVO and 1/4 cup distilled water to a quart jar. Then I shake well, and leave it settle for a few hours to seperate the water. Then I use a syringe to suck the water off the bottom, put it in another glass, and measure with the TDS meter.

I think this is an accurate, cheap and quick way to see if a mistwash is needed, and if you test the mistwash water after, if it worked. It would be interesting to see some other results with nasty WVO, I only have good sources. Now the trick is trying it on enough batches to know what the threshold of a bad TDS is.

The EPA says for drinking water 500ppm TDS is their recommended limit, so its also interesting to check your tap water, mine was 60. Its well water with the minerals calcium, magnesium, and iron being the main parts of TDS. (I had a engineer friend who did tests for EPA on farm wells run lab tests on my water a few years ago.)

From the instructions that came with the meter: TDS includes any inorganic element present other than the pure water molecule.

Lab test results
07 November 2006

Lab test results are in for the first samples I sent.

Sample 1 was the original WVO, sucked from dumpster with 100 mesh screen.
Sample 2 was 1 pass, 55-120F.
Sample 3 was heated, mistwash, 6 passes, 120-180F. I did some calcs. on the particle numbers, % reduction:

>14 micron, >10 micron
1 pass = 60.0%, 32.4%
6 pass = 98.5%, 93.7%

The >10 micron reduction looks very good, and that is the size that most people are filtering to. Although a 10 micron nominal filter won't come close to this, since they often only remove 50% > 10 microns.

I am surprised at the no reductions in <10 micron category. Maybe it was because of the mistwash which I didn't settle out, which put a lot of water in the rotor (4 oz), making the small particles not remove. Or the mistwash broke the particles down to smaller size, or sampling errors.

Interesting results for water reduction. #1= 708ppm, #2= 545ppm, #3=478ppm by Karl Fischer ASTM D6304.

It removed my suspended water, even with 1 pass, when I captured the output at shutdown. This shows the pan test is not detecting water below 545ppm. I ran 2 pan tests on each sample separated by several hours. #1 showed lots of bubbles. #2 and 3 had no bubbles. When I shutdown after 1 pass and captured the output, I tested the captured WVO and it was very bad on the pan test (lots of crackles.) This showed it was removing the water by true centrifuge in the rotor, not by evaporative means (which could make the pan test not reliable.)

Here are the # of particles per ml by HIAC HL-1185 method:
sample 1,2,3
06-10, 23483, 29230, 23438
10-14, 18067, 14542, 1560
14-25, 8559, 3482, 129
>25, 132, 58, 5

More testing is needed to determine the best way if you want to remove <10 microns (more passes, leave mistwash water settle?) My truck is a 10 micron stock filter so I am happy with the 94% reduction to 10 micron. We now have more public testing data for this than for any other filter method. I would like to see the same test run on a 2 micron absolute or 10 micron nominal filter as a comparison, but I can't find anyone who has published this data for any fluid, much less WVO.

Parts LIst

1 Dieselcraft OC-20 centrifuge from $220 (now $299 with relief and guage and tees)
1 -Power Steering Pump from junkyard $20
1 - HP or greater 110v electric motor 1725rpm (Junkyard or Harbor Freight cheapie) $20
1 -1.5” pulley wheel for the electric motor $5
1 -0.5" V-belt
Or for an off the shelf pump/motor:
pump only look at look at part# 4271K1 $223
These should work with an OC-20 or 50, and ff25 and 60:
pump only look at look at part# 4271K12 $231
pump/motor ready to go part# 4272K31 $368

Here is another source that offers both the centrifuge and pump/motor combos

1 pressure relief valve 25-250psi (grainger part 6D915) $5 or mcmaster part# 8088K14 $21
5ft Length of hydraulic hose (get it from the car you pull the PS pump from)
8ft Length of 1/2” or bigger hi-temp hose to attach to input of PS pump
1 -24” x 1.25” steel pipe (for heating element) ~$5
1 or 2-4500KW 240v water heater heating element (at 110v = 1125W) ~$8
1 -1.25” steel T (use 2 tees if you’re putting 2 heating elements in) ~$4
1 -8” x 1/2” steel tube to attach to centrifuge input ~$4
1 Reducer to attach centrifuge 1/2" to 1.25” ~$5
1 Reducer to attach PS output to 1.25” input ~$10

Plumbing fittings to attach to the bottom of the barrel to flow to the input of the PS pump ~$15
1 Pressure Gauge (grainger #5WZ34, $8.58) ~$9
1/16" thick by 1.25" ID metal washer first on the element, then a 1.5" ID washer with a 1/8" x 1.25" O-ring inside (to seal the heating element) from a Tractor parts counter

Got a new pump
26 January 2007

Several people have reported the vanes sticking in their power steering pumps, making it not work until its disassembled and cleaned. A gear pump is better since it won't have that problem. Here is my new pump/motor combo, its an Oberdorfer 991N gear pump close coupled to a 1/3hp 1725rpm motor that I got on ebay for $23.00 which is working nicely:

Added a switch panel:

I added a switch panel with extension cords and 4 switches connected to 4 outlets, which I use to turn on and off the pump and 2 heaters separately. This saves me from unplugging those 3 plugs each time I shut down. With an extra switch and outlet for future ideas.

Mobile "Holy Grail" version of this rig:

I made a mobile unit out of this same rig, to use when I am travelling long distances so I don't have to carry hundreds of gallons of Veg. Oil. I run it with a little, quiet honda 2000watt generator, or plugin power if I am stopped where I have it, which I prefer since then its easier to check and clean the centrifuge rotor. I put a sealed lid on the barrel so it can carry WVO while moving, or have the centrifuge running to process it. This is why I have the centrifuge mounted below the rim of the barrel.

With the lid, it dewaters at about 30-50% of the speed of when the lid is off the barrel.