Another Knucklehead Build

Discussion in 'A Work In Progress' started by mayhugh1, Mar 22, 2018.

Help Support HMEM by donating using the link above.
  1. Jun 30, 2018 #81

    Draw-Tech

    Draw-Tech

    Draw-Tech

    Member

    Joined:
    Jan 26, 2012
    Messages:
    151
    Likes Received:
    90
    I have been following your build, It looks like you will be the serial number 002
    I wish I had a cnc machine, you make things so easy. GREAT JOB, The crank was designed from pictures, scaling in cad, if you need any files, send me a PM
    I would like to hear from you, again "AWESOME"
    Jack
    Draw-Tech
     
  2. Jul 6, 2018 #82

    mayhugh1

    mayhugh1

    mayhugh1

    Project of the Month Winner!!! Project of the Month Winner

    Joined:
    Oct 24, 2007
    Messages:
    678
    Likes Received:
    1,999
    The crankshaft halves were sawed apart in the bandsaw, and each flywheel finished up in the lathe while being supported by its shaft in a collet chuck. Before each facing operation, the outer edge of the already finished face was indicated and the part rotated in the collet for minimum indicated wobble (for some reason it made a difference). Both halves wound up at about two tenths.

    While thinking about how I was going to offset mount the flywheels in the lathe for their taper operations (my Wabeco four jaw can't handle the long shaft sticking out the backs of the flywheels), I made a fixture that I hoped would allow me to accurately assemble the flywheels without having to hammer on them later. Basically, what I came up with was a sleeve to hold the two flywheels in alignment while the crankpin nuts were tightened.

    Although Delrin would have been a better choice, I machined the sleeve from a block of mystery plastic that I had on hand. It was faced to a thickness of 1.6" and its o.d. turned to roughly 4 inches. After boring out its center to about 2-1/4 inches, the sleeve was transferred to the bandsaw where three equally spaced 3/8" long radial cuts were sawed into its outer perimeter. On the mill, one of these relief cuts was opened up into a 3/4" wide slot, and a shim was machined to fit tightly inside it. Finally, a saw cut was made through the center of the slot to open it up to the bore.

    With the shim in place inside the slot, the workpiece was returned to the lathe chuck with the jaws centered between the relief cuts. The purpose of the cuts is allow the jaws to uniformly close the sleeve around the shim (and later around the crankshaft flywheels) as they're tightened. For safety, I ty-wrap'd the shim in place while the i.d. of the sleeve was finish-bored to the exact o.d. of the crankshaft flywheels.

    With the shim discarded, the crank halves will be assembled inside the sleeve with the connecting rods sticking up through the slot. This assembly will be done with the sleeve clamped in the three-jaw chuck while the crankpin nuts are tightened.

    In order to offset turn the flywheel tapers, I machined a carefully squared-up aluminum spacer block to stand the parts off the surface of my five inch four-jaw chuck. This spacer provided the needed clearance for the shaft, and it moved the workpiece away from the jaws so it could be easily accessed by the tiny boring bar used to turn the tapers. A concern, though, was that this block would be yet another piece in the accuracy chain. The hole for the flywheel shafts was drilled and reamed with the spacer mounted in the chuck and tight against its face. The pinch bolt machining was done later on the mill.

    Since I planned to spin the nearly 20 ounce unbalanced load in the little chuck at close to 1 krpm I was concerned about vibration. I carefully computed the requirements for a pair of counterweights that I bolted to the chuck. With these weights, the loaded chuck showed no signs of vibration even at 1500 rpm.

    In addition to indicating the pilot holes for the tapers, it was also important to indicate across the faces of the flywheels in two orthogonal directions in order to ensure the axis of the taper ended up parallel to the axis of the shaft. Any perpendicularity errors will be magnified during assembly and affect the shaft TIR's. The spacer block that I had to use complicated the setup because it tended to lift during the jaw adjustments. Protective steel shims were used between the jaws and the spacer block in order to prevent the jaws from biting into the aluminum and making it difficult to keep the block down against the chuck. After a lot of tweaking, I was able to obtain nearly identical TIR's on both flywheels: .0005" on the pilot hole, zero along one surface direction, and .0015" along the other direction. I didn't like the .0015", but it was the best I could do. After all the prep work, the two minutes of lathe work were uneventful.

    In preparation for the crankpin, I machined the nuts that will be used on its ends. Although shortened commercial nuts could probably have been used, I machined a pair from drill rod that I hardened. I wanted the finished nuts available before lathe-threading the crankpin so I could turn its threads for a close fit to them. - Terry
    225.JPG 226.JPG 227.JPG 228.JPG 229.JPG 230.JPG 231.JPG 232.JPG
     
    Last edited: Jul 6, 2018
    sbdtasos, brotherbear, kvom and 3 others like this.
  3. Jul 9, 2018 #83

    mayhugh1

    mayhugh1

    mayhugh1

    Project of the Month Winner!!! Project of the Month Winner

    Joined:
    Oct 24, 2007
    Messages:
    678
    Likes Received:
    1,999
    I ran into major problems while trying to machine the crankpin. The crankpin sets the separation between the flywheels which in turn establishes the clearance for the connecting rods and the thrust clearance of the crankshaft. The smallest of these (and it's actually pretty generous) is the specified .004" thrust clearance. In this engine with its tapered crankpin, it would have been much easier to complete the crankshaft before machining the crankcase. With the crankcase already finished, however, the clearances now depend upon the exact placement of the tapers on either end of the crankpin.

    My original plan was to CNC machine the entire crankpin in one operation on my Wabeco, and let the CAD/CAM software handle the placement of the tapers. I fully expected to have to machine a few trial parts in order to fine tune the operation, but after an entire day of head-scratching and scrapped parts, it became obvious that something was way off.

    Eventually, I recalled problems with a much earlier project in which I discovered that my CAM software doesn't seem to properly handle the profile of a lathe tool's cutting tip when turning non-cylindrical parts. I tried to characterize the error so I could compensate for it by changing the design of the pin. However, tool and/or part deflection as well as the lathe's own taper issues were combining with the runout of the live tailstock that I was using to overwhelm the part's accuracy requirements. Errors from all these sources, and probably others, were inconsistently combining to give an essentially random result on every part that I made. In addition, since I was checking the results by assembling a finished pin with the flywheels (sans nuts) and measuring the total outside width, I missed the fact that the errors were different on each end of the part.

    At this point it would have been reasonable to modify the crankcase or the flywheels to accommodate one of the already finished pins. However, measurements showed there were almost half degree differences between the taper angles on the two ends of the parts. Although the pins appeared to fit the flywheel tapers quite well, I was concerned that the angle errors might show up later as alignment inconsistencies.

    My next attempt was to machine only one end of the pin at a time. The tailstock was no longer needed, and the reduced stick-out would mean less deflection. Since I was using polished drill rod with an o.d. that matched the i.d. of the connecting rod bearings, I compiled a single program to turn the exact same taper on each end of the part. This required indicating each end of each part, since my so-called 'set-true' chuck had to be re-tweaked after every spin-up. In order to reduce the effects of the CAM error, I changed the insert to a DCMT21.50 which is a sharp diamond-shaped finishing insert with a nose radius of only .008" - half of what I had been using. I also compiled the cutting program to take only .003" (radius) d.o.c.'s.

    Two new issues arose, however. The workpiece had to be pre-turned to an exact length, and the nose of the cutting tool had to be perfectly centered over the Z=0 edge of the part. Since typical errors (for me) in either one of these steps could easily exceed the part's per-side error budget, I decided that I'd just have to make parts until I got at least one that I was happy with.

    Even though, with the help of a 10X magnifier, I became pretty good at centering the tool over the starting edge of the part, another error crept in over time that kept me off balance. The delicate tool tip wore quicker than I was accustomed to, and its tip slowly and asymmetrically changed shape. This created a continually changing error over the entire run of parts that kept me chasing my tail.
    Eventually, after more than a dozen parts, I had five with thrust clearances I could live with. I threaded all five so I could see how well the TIR's had been controlled during all the operations require to get to a final assembly. I've included a photo of a page in my notes that shows all the results. I was also anxious to see how well the alignment sleeve worked and how consistently I could reassemble the same set of parts.

    The sleeve worked as well as I had hoped. As the notes show, the final shaft TIR's were generally less than .001". As expected, there was a lot of minute structure in the runout's due to all the tiny errors accumulated from so many sources. In only one or two parts did the errors combine constructively and cause the TIR to exceed .001".

    The alignment sleeve worked very consistently and gave the same TIR results on repeated assemblies using a couple different pins. I also compared the sleeve alignment with an alignment done by tightening the nuts with the flywheels clamped in a v-block since this was essentially the assembly method used in one of the Youtube videos that I watched earlier. My best result using that method was on the order of .008" which was in line with the results obtained in the video.

    The next part will be something less critical and hopefully more fun. I'm currently planning to machine a double roller sprocket for the engine's output shaft that should add a little more rotational inertia to the crankshaft. - Terry

    233.JPG 234.JPG 235.JPG 236.JPG 237.JPG 238.JPG 239.JPG 240.JPG 241.JPG
     
    minh-thanh, bouch, sbdtasos and 4 others like this.
  4. Jul 9, 2018 #84

    vederstein

    vederstein

    vederstein

    Must do dumb things....

    Joined:
    Feb 26, 2011
    Messages:
    480
    Likes Received:
    321
    I, as many others, are amazed by the work you do.
     
  5. Jul 10, 2018 #85

    kvom

    kvom

    kvom

    Administrator Staff Member Administrator Moderator Global Moderator

    Joined:
    Jun 4, 2008
    Messages:
    3,089
    Likes Received:
    583
    Once you were turning each end separately, you could have turned the tool so that the edge was parallel to the end face. Then setting Z would mean touching the tool and subtracting the radius of the tip to get z0.

    Is your CAM error failure to properly consider the offset needed before commencing the taper?

    In any case, it all worked out well in the end.
     
  6. Jul 11, 2018 #86

    mayhugh1

    mayhugh1

    mayhugh1

    Project of the Month Winner!!! Project of the Month Winner

    Joined:
    Oct 24, 2007
    Messages:
    678
    Likes Received:
    1,999
    Kvom,
    What you suggest sounds reasonable so long as an eye is kept on the cutting direction and d.o.c., but even then there are a few issues. I made a couple sketches. The first one shows the tip oriented as the software expects. When the insert type is entered into the software, it knows about its shape and it expects its tip is located at z=0 as shown by the red centroid marker in the drawing. The software supposedly uses this tip and the shape of the insert to compute the tool path trajectories. Hopefully, for d.o.c.'s greater than the radius of the tip, it continues to properly account for the shape of the insert by including its straight cutting edges. On my particular parts, the software generated roughing passes that cut while moving toward the headstock, and finishing passes that cut while moving toward the tailstock. I had control over the directions, but I elected to allow the software to selected them.

    The second sketch shows the left straight edge of the insert rotated so it's parallel to the face of the workpiece and that edge shifted toward the headstock by an amount equal to the radius of the insert's nose. When z=0 is set, the software will assume the center of the cutting tip is located as shown by the blue centroid marker which is now located asymmetrically on the insert. When cutting toward the headstock with d.o.c.'s greater than the radius of the tip (in this case greater than .008") the software will assume it is using the left-side straight cutting edge when it really isn't. When cutting toward the tailstock with a d.o.c. greater than (in this case) about .002" it will actually be cutting on the right-side straight edge when it really isn't.

    In my experience, it's hard to set a lathe tool right at z=0 (within a thousandth or two) without facing the end of the part and immediately zeroing the Z DRO. Rotating the tool so its edge is exactly parallel with the face of the part introduces another piece of the setup that needs to be accurately done or the amount the tip is to be offset will be in error. - Terry
    242.JPG
     
    Last edited: Jul 11, 2018
    Buchanan likes this.
  7. Jul 11, 2018 #87

    kvom

    kvom

    kvom

    Administrator Staff Member Administrator Moderator Global Moderator

    Joined:
    Jun 4, 2008
    Messages:
    3,089
    Likes Received:
    583
    My CAM allows the tool shape to be rotated, and as well only does the finish pass towards the headstock in cases like this, so I can see where in your case it wouldn't work. As for Z0 being within .001, my little CNC lathe would be hard pressed to get anywhere near that.
     
  8. Jul 11, 2018 #88

    JamesDTaylorSTL

    JamesDTaylorSTL

    JamesDTaylorSTL

    "Classic American V-Twin"

    Joined:
    Apr 13, 2016
    Messages:
    2
    Likes Received:
    2
    Wow! Truly incredible work! And all the time taken to help others do a build! I never liked my EVO's heavy,boxy rocker arms so now I know how to make better ones!
    Thanks!
     
  9. Jul 13, 2018 at 3:34 AM #89

    mayhugh1

    mayhugh1

    mayhugh1

    Project of the Month Winner!!! Project of the Month Winner

    Joined:
    Oct 24, 2007
    Messages:
    678
    Likes Received:
    1,999
    With the crankshaft temporarily installed in the engine, and before starting work on the motor sprocket, I wanted to make a quick trial assembly of all the major parts machined so far. In order to install the cam box cover, the recess for the crankshaft's end bearing had to be bored. With the cover mounted in its machining fixture, I used my previously recorded coordinates for the crankshaft location to bore a flanged recess for the ball bearing. This particular hole is pretty important because it not only supports the far end of the crankshaft, but it will also be a reference for locating the rest of the cover bearings that will be machined later.

    Unfortunately, when I tried to slide the doweled cover onto the cam box which was doweled to the crankcase whose halves were doweled together, the cover wouldn't begin to go on. After a few hours of measurements, I was pretty sure that everything in the engine that could cause such an error was correct, and that I must have bored the recess for the bearing in the wrong location.

    In order to check the bearing's location, I machined a bronze blank to snugly fit the recess, and I added witness marks to both it and the cover to keep track of the blank's orientation. The crankshaft was then replaced with a length of drill rod whose end had been turned to a point so I could transfer its location to the blank. With the blank then set up in a collet chuck under my mill's spindle microscope, I discovered the recess was perfectly located in the y-direction but was off by exactly .010" in the x-direction. Evidently, I previously recorded an incorrect DRO reading - something that seems to be happening more frequently these days.

    I used the transferred mark to drill and ream a hole for the shaft and then returned what was now a bronze bearing to the cover. The fit with the actual crankshaft now seemed perfect, and so I Loctite'd the bearing in place. I left its nose a little long so it would be the first bearing to engage the crankshaft during the cover installation. Since the shaft will be just above the oil level inside the sump, the bearing will receive plenty of lubrication, but I'll also mill an oil groove on the end of the crankshaft for extra measure.

    While inside the engine, it occurred to me that the crankshaft assembly along with the crankcase oil will probably occupy 90% of the volume inside the crankcase. In addition to normal blow-by, the pistons' asymmetrical pumping action will create significant pressure pulses inside the crankcase. The crankcase is sealed except for the oil return lines coming in from the engine's top-end and a pair of oil-submerged holes connecting the sumps of the crankcase and the cam box. Since the crankcase pressure would have to overcome the oil column in order to vent into the cam box, it might instead pressurize the oil returns and cause the top-end to flood with oil.

    In order to reduce the chances of this happening, I drilled a 3/16" vent hole above the oil level connecting the crankcase with the cam box in order to increase the effective volume of the crankcase and reduce the pressure pulses. The cam box will later be vented to the atmosphere, probably with a vented oil filler cap.

    Inadequate crankcase ventilation can even prevent a engine from running. I once demonstrated this on my Hodgson 9-cylinder radial. Blocking that engine's crankcase ventilation will cause it to stall. - Terry

    243.JPG 244.JPG 245.JPG 246.JPG 247.JPG
     
    Last edited: Jul 13, 2018 at 7:18 AM
    minh-thanh, sbdtasos, kvom and 3 others like this.
  10. Jul 19, 2018 at 3:42 AM #90

    mayhugh1

    mayhugh1

    mayhugh1

    Project of the Month Winner!!! Project of the Month Winner

    Joined:
    Oct 24, 2007
    Messages:
    678
    Likes Received:
    1,999
    The Draw-Tech Knucklehead includes an external flywheel to supplement the rotational inertia of its crankshaft. Rotational inertia provides the angular momentum needed to keep the crankshaft in a one or two cylinder 4-cycle engine spinning between its relatively infrequent plug firings. Because rotational inertia scales exponentially with a rotating mass's diameter, the flywheels inside a scaled-down motorcycle engine crankcase are seldom sufficient to provide a satisfyingly low idle speed. The equation for the rotational inertia of a ring, which is the most important part of a flywheel, is:

    J = (m/2) * (R1^2 + R2^2)

    where m is the mass of the ring, R1 is its outside radius, and R2 is its inside radius. There's an additional radius dependency hidden inside the expression for the ring's mass that, when brought out, shows the rotational inertia is an extremely sensitive function of the ring's diameter:

    J = (pi/2) * t * D * (R1^4 - R2^4)

    where t is the thickness of the ring, and D is the density of the metal (approximately 4.5 oz/cubic inch for brass or steel), and the sign change isn't a typo. This form of the equation is seldom found in physics texts, but it's most useful for design work. The rotational inertia of a simple disk can be found by setting R2=0.

    Rotational inertia is easy to compute, but the amount needed by a particular model engine for a well-behaved idle isn't so simple to estimate. In order to test the need for the Knucklehead's huge five inch flywheel, I looked at a couple popular model engines that are similar in design and scale to the Knucklehead and known to behave at low idle speeds.

    The first was Jerry Howell's 90 degree V-twin which was actually my first IC engine built some eight years ago. I calculated the total rotational inertia of its entire rotating mass to be approximately 97 oz-inch^2 with 90% of it coming from its (4"x 1-1/8" thick) external flywheel. The second engine was the 45 degree Hoglet whose plans appeared in Model Engine Builder magazine. Remarkably, at 105 oz-inch^2, its rotational inertia is nearly identical to that of Jerry's engine. Almost all of it comes from the two 4" x 9/16" thick flywheel rings integrated into the crankshaft inside its skeleton crankcase.

    The rotational inertia of the pair of internal crankshaft flywheels inside the Draw-Tech Knucklehead is a healthy 26 oz-inch^2, and its external (5" x 3/4") flywheel adds another 127 oz-inch^2. Its large external brass flywheel is a clever multipurpose design that includes a rope start pulley as well as some limited cooling from its fin-shaped spokes. For my taste, though, it's too large and doesn't look at home on a motorcycle engine. I set a goal of 75 oz-in^2 for a new design since this would bring the total up to match the two more common v-twins. Excessive rotational inertia adds stress to the starting system and increases the torque requirement of the starter motor.

    I began by juggling the paper design dimensions of a plain steel disk that would achieve 75 oz-in^2. It quickly became apparent that even with a 1-1/4" thickness, I was going to be stuck with a diameter on the order of 3.5". For a motor pulley this was way out of whack with the rest of the engine's scale.

    I looked at shelling a portion of the disk and filling it with ultra-fine tungsten powder that I had on hand. Tungsten has a theoretical density of 10.3 oz/cubic inch compared with 4.5 oz/cubic inch for steel and would seem like a great candidate for reducing the flywheel diameter by some 23%. Unfortunately, this theoretical density can only be approached by casting tungsten which is impractical due to its extremely high melting temperature. I experimented with pressure sintering it, but the best density I could achieve using my hydraulic press was approximately 6.3 oz/cubic inch. This was an improvement over steel, but essentially identical to what would be expected with cast lead. Since the resulting reduction in flywheel diameter would be only about 8% (the fourth root of the density ratio), the effort involved in working with the messy and extremely abrasive stuff was hardly worthwhile.

    After finally accepting the fact that I wasn't going to end up with a flywheel on the output shaft disguised as a reasonably scaled motor pulley, I thought I'd at least try to dress it up a bit. I did two partial CAD designs - one for a triple roller chain 'ripper' sprocket and another for a cogged belt drive pulley. Variations of these can be found under the primary covers of Harley motorcycles.

    The sprocket was functionally designed for three 1/4" roller chains so I would have the option of using it to drive something or perhaps even to start the engine. The teeth added much more to the diameter of the sprocket than they did to its rotational inertia, and so its interior would have to be shelled and filled with lead in order to keep its diameter below that of the crankcase. The belt drive pulley was also designed to be functional and is identical, except for the number of teeth, to one that I added to my Howell V-twin so it could drive a faux transmission.

    After a lot frustration, I decided to stay with a simple polished stainless steel flywheel that wasn't pretending to be something else. Although the more interesting versions would have been fun to machine, they attract attention to themselves and fact that they're wildly out of scale. A plain wobble-free polished flywheel, while spinning very close to the crankcase, might tend to fool one's eye into thinking it was actually part of a primary.

    The design I finally chose to machine is shown in the last photo. At 3.4" diameter and 1.25" thick, the chunk of solid stainless steel has a total rotational inertial is 72 oz-in^2. A tapered lock bushing will secure it to the output shaft, and an integrated 1" hex will provide for drill starting engine just in case I'm unable come up with a functional electric starter. - Terry

    248.JPG 249.JPG 250.JPG 251.JPG 252.JPG 253.JPG
     
    Scott_M likes this.
  11. Jul 19, 2018 at 4:45 AM #91

    GlennS

    GlennS

    GlennS

    Member

    Joined:
    Aug 19, 2017
    Messages:
    5
    Likes Received:
    0
    Question: if you were to add the primary chain and driven (transmission input) pulley, would their inertial weight become part of the flywheel equation?
     
  12. Jul 19, 2018 at 8:14 AM #92

    mayhugh1

    mayhugh1

    mayhugh1

    Project of the Month Winner!!! Project of the Month Winner

    Joined:
    Oct 24, 2007
    Messages:
    678
    Likes Received:
    1,999
    Glenn,
    Yes, any inertia in the load would be included. The chain would also add some frictional losses that the engine would have to overcome.
    Terry
     

Share This Page