Oscillator Cylinder Pattern

[vtsteam]

Thanks CI, Jason, Paul! 

I consider myself about 25% lucky to get a good iron casting, and 75% sure, due to experience gained from a long line of former failures! 

I used a wider than necessary chucking extension on the castings to provide material for the long pushrod support, and the piston nut recess on the other cover. Paul, The bottom end of the hole is unthreaded for a packing space, and the bottom was left as drilled, not flattened. I was thinking of trying to taper it even more, but I don't know if that's necessary. Jason, probably knows. I was planning on trying some kind of wound packing, not an O-ring.

[Jasonb]

I usually just leave them as drilled and use graphite yarn.

[vtsteam]

Okay, ordered some graphite yarn. While I was at it, I went down a rabbit hole trying to decide what size holes to drill for the fasteners in the cylinder flanges and covers. I probably spent two hours going back and forth to the shop, looking online, etc.

1/4"-20 cap bolts just seemed too large in every way, but #10-24 was the only next size down I have, and those seemed too small for strength. #12-24 seemed about right, but they are, these days, not a normal size for hex head cap bolts, and in fact for little else as well.

I thought about metric sizes (I now have new taps and dies for them) but 5mm was still small seeming, and 6mm pretty much a 1/4 bolt. #12 still seemed like the right size.

Looking at what was available in 12-24 from the big river outfit, it was mostly socket head cap screws, which would have been okay functionally, but I just would have liked hex head better. Then I thought about studs and nuts -- were those available in #12. Well not really for the all-thread, and the nuts were very expensive and 100 quantity. I don't think I'll have a big need for them in future.

Finally I hit on the idea of buying an assortment package of #12-24 Phillips pan head 304 stainless screws, with washers and nuts. I can saw off the heads of the longer screws to make studs. Nuts are included already, and the shorter screws can be used as temporary fasteners before I get too fancy with the studs. The whole package of 105 pieces cost a little more than half of what 20 socket head cap screws would have. So a solution of sorts.

[crueby]

Making up your own studs like that is a great way to go. I like to do that, using high strength red loctite to hold the nuts in place on the threaded section. Depending how long the screw is, you can get more than one stud per screw. Let them set up a bit then wipe off any excess loctite outside the nut.

[Jasonb]

For the small amount of nuts you are going to need milling them from round stock is not going to take long. The advantage is you can go a bit smaller with the hex size as I find stock UNC nuts rather large on the A/F and flat.

As for size/strength a lot will depend on how many fixings you are going to use 4, 6 or 8?

By way of comparrison a Stuart No4 or No9 which are also 1 1/2" bore and easily capable of taking 50psi have just four 4BA fixings which are just a couple of thou larger than a #6 UNC.

[vtsteam]

Thanks guys! 

Jason, wow a little bigger than #6! That makes the #10's I already have seem overkill. Well now, maybe I'll wait on drilling, just to see how the new fasteners look when I get them (tomorrow, according to the claim). I was going to do the holes and threading today. I had been thinking of using 6 studs per cap. But I'll hold off.

My thoughts on making a boiler were for a monotube type, with an intended modest working pressure of 50psi, but it seemed to me I ought to err on the side of strength in the steam cylinder construction, anyway.
 

[vtsteam]

The screws didn't arrive today, so I didn't want to drill the end caps. Instead I hunted up a cast iron riser from one of the pours, and roughed out a piston from it. I first turned it to rough dimensions, then center drilled it, drilled for tapping, and tapped 5/16" -18. Finally I parted it off.

I've got some 5/16 drill rod that I'll use for the piston rod, but that was as far as I got today. I did spend some time thinking more about the flywheel pattern. I have a few options, one of which would be doing most everything with hand tools, which I'm confident about ... and the opposite -- resurrecting my cnc router, and seeing how much wood I can ruin that way.

I have never used it for 3D profiling, only 2D cutouts in aluminum, so there would be a definite learning curve. I can generate STL files from sketchup, but CAM is another matter. Still thinking on it.....

[Jasonb]

.STP would probably be a better file format and is what I use. What CAM do you use?

You should end up with two halves that will be almost reasy to paint just needing a quick sand depending on teh wood chosen, tighter the grain the better so if you have it maple, cherry and even poplar if the tools are sharp. Rough out with an adaptive then a path that scallops rather than ramps or contours will give the best finish with a ball nosed tool.

[vtsteam]

Thanks for the ideas, Jason. I looked into it further last evening. I have only used the SketchuCam add-on with SketchUp 8 (the old Google free version) That's only a 2.5D cam addon. I do have an STL output addon for that old SketchUp version, and did a couple times successfully run its STL file output through Cura slicer, to produce a 3dprint.

I use Linux, and run LinuxCNC (old version) to work with the router.

I watched this and this guy's related flwheel videos last night.

 <a href="https://www.youtube.com/watch?v=31pdXyVm3hg" target="_blank">http://www.youtube.com/watch?v=31pdXyVm3hg</a>

And gave a lot of thought to whether I want to go this route rather than simply carve out a pattern. The running of a machine in an unheated shed for 2-1/2 hours in below freezing weather while standing around watching it is not something I want to do, and I'm not comfortable just leaving it on its own, if past experience with issues is any guide. And that 2-1/2 hours is only half the pattern. The location of the shed where the router is located is 300 feet down hill from my house over what is presently ice covered dirt drive. I can't just look in every 10 minutes.

What I'm favoring now is a hybrid route -- have the machine make a kit of parts that I can assemble into the flywheel: a hub, a rim, and spokes (all in halves for a split pattern). This would cut the jobs down to reasonable chunks of time, and I believe the toolpaths could be optimized better for each shape, and with less waste cutout time. I also favor doing fillets by hand on the pattern, rather than trying to program my limited CAD 3d program to do that in the most complex areas -- though I won't know until I try how far I can get, with those -- we'll see.

I did find a free version CAM program that works in Linux and accepts STL files that I think I'll try called DeskProto. In fact I did try it at home, manipulating an object for cutting on a quickly drawn spoke to see how it works, and I think it will be adequate for doing what I want with this flywheel.

So that's the tentative plan, a half cnc, half human job of it which suits my quirks, and shop situation, in the cold.

[Jasonb]

I have done similar on larger flywheels doing the spoke and a segment of the hub as one part cut multiple times. Hopefully if you watch the first one and it runs OK you are less likely to need to watch the remaining ones.

Rim could be done on the lathe or a rotary table on the mill and then just glue it all together.

I include all the fillets as the CAM works out all the paths whether it's a curved spoke or curved fillet does not matter to the software.

Don't forget if doing curved spokes you need a right and left hand set!

[vtsteam]

Nice spokes Jason.

I'm not planning the S-curvy ones, just tapered eliptical straight ones. After playing around with shapes in CAD and the free version of Protodesk CAM, I think it's going to be wayyyyyyy too slow at toolpath generation to ever consider doing the entire flywheel. In fact when I asked it to do an 8" diameter square cornered simple ring, 1/2: thick with a 1/8" mill, NO spokes, no hub,  it said it would take hours to machine. I'm sure the paid versions have many more toolpath varieties than the single one in the free version, so must be faster.

But it did say that a single split tapered elliptical sectioned spoke would take only 12 minutes (or 6, or 3 if I want a progressively coarser version). I'd need 12 of them. Maybe I'll try that. On the other hand that's probably 2 total hours by the time you're done with it, and I'd also have to provide wood of quarter inch thickness not to have to do roughing passes (which the free version doesn't seem to do also) so there's additional manual work in preparing stock, positioning it, etc.

If I have to prepare thicknessed blanks for CNC stock via table saw, I might just do an end run and try whittling out a spoke from a sized blank by hand and see how long that takes me. That can be done in the house where it's warm. As a boatbuilder I used to make tapered masts and spars with only a hand made gauge a drawknife and plane ... heck, these spokes are the size of clothes pins!

[Jasonb]

Not unknown to make one pattern and cast 12 spoke halves.

[CI]

Or cast the whole spoke and use a follower.
You can mix up some sodium silicate sand, fill a flask half with that sand, cover with a sheet of kitchen plastic wrap, press the full pattern half way into the sand, and let the sand set.

The follower sand does not have to be exact or perfect; it just has to support the pattern from below when a sand mold is rammed.
I have also used bondo to make a follower in a similar fashion, but for larger followers, sand with sodium silicate is less expensive.
The sand can be low grade, for the follower.
And lastly, I have used plaster to make a follower, but sodium silicate bound sand is the easiest and lest expensive method to make a follower.

A solid spoke is much stronger than half a spoke.
I had some half-spoke patterns, and then tended to break at the thinnest point of the spoke.

All of the Cretors patterns, including the flywheel patterns, were one-piece; no spit patterns.
.

[vtsteam]

CI I was planning on a pattern plate. I could reduce the number of half spokes needed by making that single-sided. Below is a pattern plate I did for a Westinghouse type twin.

Been waiting two days for deliveries supposedly due to ice on drive, even though sanded twice, and big metal delivery box is near entrance. I re-sanded the entrance an hour ago just to make it obvious it's no problem -- we've been driving in and out for days with cars and pickup truck. Must be a new driver. 

In those deliveries are the fasteners for the cylinder covers, graphite yarn, a new 5/16" die and handle, and two ball-ended mills -- so I'm on hold for most things needed to continue. 

[vtsteam]

Wow success on all fronts!

When the deliveries hadn't arrived this afternoon, I got tired of waiting, and decided to just try a CNC cut with what I had on hand, which meant a 1/8 square nosed cutter used for aluminum, an old discolored piece of 1/4" birch plywood for material, and a CNC file of the spoke generated by DeskProto at a low resolution. I didn't expect much. Maybe a ball of fuzzy plywood splinters from the core or a burning smell, and at best a rough approximation my oval spoke

As a second thought after Deskproto had created the cut file, I went into the G-code and changed the default feed rate from 40 IPM (the default) to 20 IPM with a text editor. I didn't want to break the bit -- an Amana single flute up cut. I ran my spindle at it's highest speed and just figured I'd check for burning and slow it down if necessary.

I stuck the ply down with some double sided tape (which I figured didn't have a big chance of holding at the 35 degree shop temp, but the hell with it, I want action!) I zeroed that axes. and then just let her rip.

Well heck, everything behaved perfectly. The piece was machined without issue. After I removed it and dusted it off, it seemed it was even useable  as a part! The plies were of good quality wood so no splintering and just slight fuzzies on the climb milling side, which could be easily sanded off. So, okay what about another? Well why not...I just jogged the X over, re-zeroed that axis and hit go again. And it cut another spoke half!

By then I was happy enough to want to quit, and take a closer look at the pieces in the house. But first might as well check the package box at the head of the driveway. ...Well, whadya know, all my late packages (5 total) were delivered, too! 

Well, seems silly to get excited over a couple little sticks of wood, but I admit that I am, and here they are, two halves that I'm holding together, as machined, without any sanding or cleanup. Plywood, no less. I have on hand cherry, birch, basswood lumber cut from trees on my property, much better for carving, but surprisingly, the plywood test material looks really quite usable out of this router.

I've also just realized that tho the DeskProto CAM free edition doesn't have a good toolpath for the rim, (supposedly it would take hours  ) my onboard SketchUp add-on is a 2.5D capable g-code generator, and able to program it and provide a fast toolpath for that kind of shape. So really I can do the pattern, I think, as a kit of parts from the CNC router.