An Article submitted to "The Funnel", The journal of the
SBA of Great Britain
Some thoughts as to
construction methods, some interesting calculations, a method of fabrication
introducing a way to reduce cumulative error, a failure, and a design for a
There have been very few
articles in "The Funnel" or "Steamboating" or any of my steam-related
literature on how to make your own crankshaft. Perhaps that should have
warned me before I even started.
The finished shaft prior to
grinding, all 32" of it
The following account relates
my experience to date.
October 2001 found me seven
months into the machining of a 2¾ + 4⅜ + 6½ x 4" triple expansion marine
steam engine. The castings had been purchased from the Elliot Bay Steam
Launch Co. in Portland, Oregon.
The castings for the engine
are supplied on a "not ready for production" basis which essentially means
that if you machine to the drawings it will not all go together.
(See Builders Notes)
I had taken a trip to the U.S.
in July 2000 to speak with an engine builder in Colorado who is about 70%
finished and to another in Vancouver. The latter finished the engine a few
years ago and it runs magnificently in a 23 foot Elliot Bay hull. Convinced
that it could be done I bought the castings and they arrived in October
The workmanship of both of
these men is exceptional and both have been a great help to my amateur, self
- taught efforts so far. Another of these engines has just been completed in
Maine and five are currently under construction in the UK.
By October I had completed the
cylinder block, bedplate, main bearings, pistons, slidebars, crossheads,
connecting rods etc. etc. Progress to that date is shown in
an adjacent photo.
|Progress at October 2001
I had, for some time, been
losing a bit of sleep thinking about the crankshaft.
A casting for the 32" long
crankshaft is not currently available and the options and conclusions
appeared to be:
by shrink-fitting: The traditional method for large shafts (ship sized), a
bit stressful, thought I, with the chance of a journal locking up half way
through a very hot web and wrecking the whole job. The method used in the
pattern and cast it: The traditional method for smaller shafts but
although the pattern would
not be a big problem, finding someone to actually cast one in the correct
material and the large lathe (11" swing over the saddle) required to
machine the crankpins ruled this method out for me.
from the solid: Not too bad for small shafts but machining all that metal
from 11" diameter stock and the large lathe required ruled this out
blank out of flat stock, twist it to provide the throws and then machine:
A novel method being used in Australia for a couple of compound engines
but not so far attempted for three throws.
and weld/ braze: The method used by the builder of the five UK engines,
distortion likely to be a problem, annealing and grinding necessary.
by machining webs, pins and journals then "locking" together by splitting
the webs and securing by squeezing with socket head screws: The method
used for the two US engines, thought to be suitable for up to 20HP. No
grinding thought to be required as the shaft is assembled and locked in
position in the bearings. I had some doubts as to torsional strength.
and secure with Loctite™: Certainly a low machinist-stress option,
strength assured say the technical representatives but what happens if you
run a hot bearing and the adhesive fails above 150 degrees C?
I eventually chose the shrink
fit after speaking with the Canadian engine builder but decided to eliminate
the keying of the journals and webs that he used as I had doubts about my
ability to machine them accurately. The impression that I had was that the
keyways were primarily there for the purpose of location rather than
torsional strength in a shrink fit joint. I would be interested to hear the
opinions of others on this matter.
diameter of the final shaft was to be 1.500" so I used 1.625" to allow for
final machining to eliminate any eccentricity, and grinding. The dimension
of the webs was 2.500" wide, 0.875" thick. With the journals and pins in the
webs there was 0.750" of material beyond the main journals and 0.450" of
material beyond the crankpins. Obviously, there was 0.500" of material on
either side of the pins and journals. Drawing 1 shows the essential
These dimensions were taken
from the drawings supplied with the engine, which were designed for method
five listed above. My assumption that these dimensions were suitable for an
alternative method of construction led to a problem explained later.
My first shrink-fit experiment
went surprisingly well, I wondered what all the fuss was about. The web was
bored 1.4935", the pin 1.495". Heated with an LPG torch until the hole
measured 1.502" the pin fell into the web and after cooling I was unable to
hold the shaft firmly enough to see how strong it was.
My second experiment was to
make up the components for a single throw shaft to see if my assembly
technique would work: crankpin into hot web, web and pin slid into hot web
on a flat surface to produce a "web assembly" and the main journals dropped
into the remaining holes. I was a bit careless with measuring the hole in
the hot web for the first journal and experienced my first "lock up". The
journal went in at an angle, locked halfway and that was that. My confidence
By cutting off the journal
close to the web I was able to drill a 1.250" hole axially through the
offending journal and was able to knock the remainder out with ease leaving
the web intact. Although a bit upset about the failure, I had
way that I could get out of any future trouble. Previously I had nightmares
about dropping the last journal into place on the almost complete shaft,
having a lock up, and having to start all over again.
The first hint of possible
trouble occurred when I put the trial crankshaft into the dividing head on
the mill, clamped a substantial bar to the outermost web and leant on it.
The crankpin joint failed with very little effort. This is the joint with
0.450" of web outside it. I was not able to hold the main journal tightly
enough to break its grip on the web. Puzzled, I attributed the failure to a
poor finish in the holes in the web leaving less surface area to grip.
I made a further trial "web
pair" and clamped one of the webs to the six foot leg of my workshop crane.
I clamped an eight foot length of 1.5" square tube to the other web and
gradually leaned on the end. The joint failed just before my 150lbs of
weight was fully supported on the tube…. I estimated 1000 foot-pounds of
turning effort on the joint.
|The test rig, less the eight foot
When I inserted two 0.250"
silver steel pins through the joints I was able to stand on the end of the
tube before the joint failed… I estimated 1200 foot-pounds (150lbs at eight
feet) Note that this trial was carried out on the main journal end of a web
ie. with 0.750" of material beyond the journal. Also on this trial piece
there was only one hole in each web
It was only then that I
calculated the maximum torque likely to be produced by the engine.
Each of the three engines is
likely to impart a turning moment of a maximum of around 180 foot-pounds (eg
180psi on the 2.750 diameter HP piston gives a "push" of 1062lbs at a
distance of two inches. This equates to 177 foot-lbs when the crankpin is 90
degrees past top dead centre. (2/12 x 1062))
With a three cylinder engine,
the maximum turning effort appears to occur when one of the cranks is at 120
degrees past TDC and another is 120 degrees before TDC. The other crank is
on TDC and cannot generate any turning force. Because the crank is no longer
horizontal it's distance from centre is reduced to about 1.750" and
therefore the torque at that point is approximately 150 foot-pounds for each
of those two cylinders (1.750/12 x 1062).
300 foot-pounds was just a
number to me until I realized that the 5 litre 300HP V8 in my wife's
four-wheel drive produces similar numbers.
I knew that steam engines
produce a lot of torque but it still seems a bit hard to believe that my
engine designed for 10 IHP @ 300RPM is likely to produce that much twist!
That figure of 300 foot-pounds
is for maximum engine torque. To allow for shock loading such as hitting a
rock with the propeller or getting a slug of water in the engine I
understand that double that amount is considered a reasonable figure to work
with. So with a joint thought to be safe to 1200foot-lbs, a two to one
safety factor seemed a bit tight but probably safe.
Back to the shaft:
The three pairs of webs were
machined then surface ground to final size, the holes for the pins and
journals were bored to 1.492" using a substantial boring bar and a tungsten
tipped tool to get an excellent finish. The three crankpins were machined to
length then the ends were turned to 1.494" and the journal surfaces finished
to 1.560" to allow plenty of meat for grinding out any distortion or
The five main journals were
machined to length with one end of each turned to 1.494".
Using the LPG torch and a
crayon that changes colour when a certain temperature was reached. I heated
up each of the webs to 600 degrees F. At that temperature the bores measured
|Assembling the "web pairs"
I assembled the three web
pairs as previously described. Into the LP pair I inserted both main
journals, into the IP and HP web pairs I inserted only the forward main
Ready to insert a journal into a web pair
When I placed the aftermost LP
journal/web pair/journal assembly in the lathe the runout of the forward
journal was about 0.015". I centred this journal and turned it to 1.520" and
reduced the exposed end to the shrink fit size.
This procedure "corrected" the
assembly misalignment so that any error did not accumulate down the length
of the shaft.
With this assembly in the
dividing head on the mill table and the crank positioned horizontally with a
dial gauge I wound it round 120 degrees (The LP trails the IP). With the IP
web/ forward journal assembly set up on parallels at centre height I heated
up the aft IP web and slid it along the parallels and onto the forward LP
|Assembling the crankshaft in correct
angular alignment on the mill table
This IP forward journal of the
assembly was "corrected" in the lathe as previously described but using a
steady on the IP aft journal.
"Correcting" the crankshaft after assembling the IP
The same procedure was used to
locate the HP web pair/journal assembly, the crankhaft now being complete
the forward main journal was "corrected" and I had a three throw crankshaft.
Seven hours on the crankshaft
grinder gave me the finished shaft. This time would have been dramatically
reduced had I not been so pessimistic about distortion and left the
crankpins at 1.520" rather than 1.560".
It looked beautiful and I
proceeded to machine the counterweights, turning them from 6 1/2" stock,
cutting them in half and milling out the recess. I fitted them all to the
webs as a gentle tap fit.
|The finished shaft
It was when I was
counterboring the attachment holes in the counterweights that I noticed that
disaster had struck... The shaft was not running true, to the tune of about
0.040" The crankpins were not absolutely tight in the webs. My tapping on of
the counterweights had upset the alignment.
To say I felt sick would be an
After tapping the shaft back
to truth (well, within 0.0015") and pinning the joints with the silver steel
pins I could still get movement in the crankpins which lets the shaft run
out 0.005"...clearly unacceptable
So, what went wrong?
When I went back to my
Machinery's Handbook on shrink fits, where to look for some of the answer
thickness or amount of metal around the hole is the most important factor".
When I measured the width of
the webs (they were surface ground to 2.500") the end with the crankpins all
measured 1.501 - 1.5015"
I originally thought that
there was not enough metal outside the crankpins (about 0.450") to stop the
web stretching around the journals. Thus, my shrink fit allowance of 0.0015
- 0.002" was reduced to a maximum of 0.001" and a minimum of 0.0005",
clearly not enough to lock the pin. The other end, the main journal end that
I tested to destruction obviously had enough (about 0.750")
Some calculations with the
assistance of an engineering design consultant in Sydney,
an old family friend,
showed that my joints were
placing a load of somewhere well in excess of 50,000lbs on the steel. Mild
steel's yield strength is about 20,000lbs.
So it appears that the major
contributing factor to the problem was my choice of mild steel for the webs.
When I had spoken to the Canadian engine builder about his technique and he
told me that he had used "4140", an alloy steel, I had assumed that this was
for the shafting, not the webs.
My understanding now is that
the 0.002" interference has to go somewhere (the journal is essentially
incompressable) and the "bulge in the webs must happen. However I had taken
the steel beyond its elastic limit , its ability to return to its original
shape and was left with a hole not small enough to grip the journal
Although I considered welding
the pins to the webs the only proper way to fix the problem was to start
I have decided to make three
changes to my original design to strengthen the joint:
webs wider (3") and longer over the crankpins (I can just get 0.750"). This
change in size will increase the amount of metal around the crankpin bore by
machine the webs from 4140, an alloy steel with a yield strength of around
reduce my shrink fit allowance to 0.0015 - 0.0018". This significantly
reduces the stress on the steel and should still give sufficient grip to the
The area of the web between
the crankpin and the journal is still the area of greatest stress and the
new design does not improve this situation.
||The design of the first crank web and
the modified MkII version
There is nothing like a test
to confirm a theory so I will assemble a single throw shaft to the new
design and stand on the end of my eight foot lever and see what happens. I
hope that the joint will support a few of my friends as well!
Lessons I hope to have
I will not
assume that a design for one method of manufacture will suit another.
check the materials specification.
I will do
all of my calculations first.
When I run
into a problem, I will think again, right from scratch.
When I test
a design I will make sure that what I am testing is exactly the same as the
I will not
despair over six week's lost work in a two-year project!
The Successful Result?
As I write this I am just
finishing the machining of the "trial" shaft to the new design.
As the deadline for the next
issue rapidly approaches you will have to wait until then to hear how strong
the new design is.
Crankshaft Update 25th April 2002
In early February I constructed
a single throw crankshaft based on the new design using “Holdax”, the 4140
alloy equivalent recommended to me. Using 0.0025” interference I could
bounce up and down on the end of the eight foot long test bar with no hint
of movement in the joint. That’s 1200 foot pounds of turning force. Looking
around for additional weight to find the joint’s final strength I suspended
my dividing head on the end of the bar as well. It weighs an estimated
75lbs. When I stood on the bar the joint failed just as my full weight was
about to be supported. I estimate that at least 200lbs was supported before
failure occurred. That’s 1600 foot pounds or a safety factor of five and a
bit for the joint.
I should mention that I had been
experimenting with different interferences and that 0.0015” failed at less
than 1200 foot pounds. The final test using 0.0025” would probably have
exceeded the 1600 foot pounds achieved if I had not experienced a “lock up”
during insertion of the pin into the web so that it was only inserted about
3/4 of the way in.
With a successful design I was
confident to proceed with the construction of the Mk II crankshaft.
Fabrication proceeded smoothly with a bit more runout at each stage of
assembly than I had experienced with the first shaft. The worst offender was
the forward HP journal that was running out 0.050” before “correction”. This
was due to a mysterious gremlin that permitted the HP web pair to shrink
with a twist. “Correction” in the lathe as previously described sorted this
out and the shaft was ground successfully to a finished size of 1.4995”.
The two outermost main bearings
were bored in a fixture in the lathe and the three inner mains were line
bored in the engine. This was accomplished on a fixture on the front of the
mill table using a flat belt drive from the mill spindle that had been
turned horizontally. Thrust was taken from either end of the boring bar with
two pieces of 1/4” silver steel propped against a wall and a vice. Feed was
applied using the table power long feed.
At this date the big ends have
been machined and fitted and I have reciprocating motion in the engine!
Despite being a little on the
tight side an electric motor coupled to the end of the crankshaft with a
large reduction turns the engine at about 200 rpm very smoothly.
Seeing the engine turn over for
the first time was more than adequate compensation for the design and
testing hiccoughs taken to get there.
My earlier comment that the
engine would not go together if machined to the drawings could be taken as a
criticism. On the contrary, the design and problem solving necessary to get
this far have been one of the most rewarding aspects of the entire process.
Not always appreciated at the time but the result is very satisfying.
||The MkI shaft and the MkII webs
||The MkII components
shaft in the engine
crankshaft and counterweights
||"Trueing" the counterweight
The Project | Builders Notes |
Workshop | Photo Gallery |
Specific Components |
Problems | Movies | Resources
Other Builders | Who Am I?