E.P. Allis 3000hp Quadruple Expansion Corliss Engine, 1893 in 1/12 scale

[Michael S.]

The component is larger than I first thought. The disc says 1 1/2 inches, and for us, 1 = one. But it's definitely seven and a half inches!
Great work.

Michael

[Chipswitheverything]

A very impressive start has been made on this formidable engine project, which in engineering terms is rather like the sort of massive model engineering usually associated with one third, or even half size, traction engines and road locomotives!  Looking forward to seeing more details of your build of this mighty steam engine as time goes on.  Dave

[bananarchy]

Flywheel Spokes - Casting Time!

One of the key factors that makes this project remotely doable is the fact that my machine shop shares space with an (amateur) iron foundry. I'd never done any pattern making or casting prior to this, so some learning curve is to be expected. The advent of 3d printing also makes this SO much easier - the ability to simply 3D print one's patterns instead of crafting things by hand is also critical to being able to make accurate parts in a timely manner. A bit of bondo and filler primer is required to smooth things out

The design and geometry of the flywheel rim segments make that pattern significantly more complex, so I started out with the much simpler flywheel spokes. The first pattern was for a single spoke, angled to fit within our most common size of casting flask. This is rammed up, and the sprue is cut by hand into the upper part of the mold for the molten iron to be poured into. The first casting turned out beautifully! Good fill, negligible shrinkage, and great surface finish. In my excitement to start machining it, I clamped it to the mill table a bit too enthusiastically and promptly broke the casting. The pattern also got broken when trying to make a second mold, due to too much air pressure in the mold squeezer. WELP. A two-cavity mold will be more efficient anyway. On to version two!

The geometry for the two-up pattern was basically unchanged, and required the use of a longer flask. The sprue goes into the runner between the two parts. The P shaped geometry is a spin trap, which is intended to fill up first and catch any loose sand or debris before the part starts to fill. Two molds were rammed and poured - the first pair had some shrinkage, so the second had risers cut into the mold. These are larger volumes of metal above the part which are intended to solidify last, and contribute material to the part as it cools and solidifies to minimize sinks and voids. Four parts!

Time for some test machining!

[bananarchy]

Flywheel Spoke - test part machining

The spoke with the worst sinks was selected as the setup part to prove out the design and machining workflow. This consists of first milling and then grinding the ends to the correct thickness to interface with the hub and rim segments, drilling the bolt holes, and then milling the exterior sides of the ends.

For the milling, the part is clamped to the mill table with two clamps across the spoke snugged down only lightly to keep it from flying off. Additional constraint is provided by toe clamps against the outside to prevent movement on the table. The machining allowances are pretty generous, since I found out rapidly that light cuts on the skins of the casting were quite unkind to the milling cutters. Both ends are milled and ground to the finished dimension of the root (1 inch) and then the head is milled and ground further to its dimension (0.750).

I realized somewhere in this process that I'd made a bit of an error in the spoke design/pattern: effectively, the head of the spoke is too big in a couple of dimensions. When getting it down to thickness, the machined area extends too far down the spoke towards the center of the flywheel, leading to the machined area extending beyond the interface with the rim segment, and being visible. This is actually the case with the root as well - the parts ended up slightly thicker than intended, and the large radius fillets leading to the pads at the ends make this a really sensitive dimension. I finished out the machining in order to learn what I could, but we're gonna need to alter the pattern and make more parts to address this.

On to drilling! Now we've got to start paying attention - how do we locate the bolt holes relative to the casting? A guide was 3D printed to show the hole locations and the finished outer dimensions of the spoke pads (as well as the surfaces where the hub and rim segments stop). After the centerlines of the spoke ends were squared up to the mill, the positioning guide is positioned and then a center punch marks the inner bolt hole. Pick up on the hole with a wiggler, then drill and ream all holes 0.251.

I whipped up an aluminum fixture plate to aid in the milling operation - since we want to locate the milled features relative to the bolt holes, this allows the part to be bolted down parallel to the X axis for the straight cuts and at an angle for the angled faces, and the cutting is done at known locations relative to the bottom right corner of the fixture plate. Shoulder screws are used in two of the holes to provide accurate location, with regular 1/4-20s in the rest of the holes for additional securement.

[CI]

Nice pattern and casting work !
Great surface finish.
Rather convenient to have an iron foundry on hand; don't see many of those.
Watching with much interest.
 

[bananarchy]

Flywheel Spoke - Test Part Cont.

The part is test fit with the completed flywheel hub. When grinding the spoke root thickness, I ended up with about .0005" clearance in the hub slot. Tighter than it probably needs to be, but very nice to feel when assembling. With the bolt holes reamed .251 in both parts, 1/4" fasteners go through the holes readily. The material for the studs I will make is about .250 on the nose, so I'll probably ream most of the holes in the final spokes .253 to give things a little more wiggle room, but it's pleasing to know that my hole locations and the squareness of the assembly are that accurate.

In the last photo, the spoke is joined with a 3d printed flywheel rim segment, and we can see that the machined face of the spoke head extends beyond where it interfaces with the rim. This is exaggerated in this instance because I intentionally biased the location fixture to make the hub interface line up where I wanted it on the casting, rather than splitting the difference. Still though, the spoke shaft needs to be thinner and the fillets need to be altered. New pattern time.

[bananarchy]

Flywheel Rim Segment

This component is a greater challenge than the flywheel spoke - owing to the angle of the ends and the shape of the interior geometry, it cannot be made with a simple two-part mold. Instead, a core, or a sand structure which is formed separately from the main mold and assembled with it, is used to form the part geometry which would otherwise be an overhang. The core is made by ramming sand into a 3D printed corebox, which another little design puzzle in itself. This whole pattern designing endeavor was a serious spatial-reasoning head-scratcher at multiple levels of remove, but I feel like I came up with a decent solution.

The second picture shows the as-cast part. The red surfaces of the part are formed by the pattern, and the grey surfaces are made by the core. The third is the pattern itself - the black part forms a hole in the mold for the core to drop into.

Next we have the 3D printed and finished casting pattern followed by the corebox, which is a 4 piece assembly.

Time to ram up a mold! Unlike the spokes, the pattern is not mirrored on both sides of the match board - the other mold half is just flat except for the sprue. Next the corebox is rammed with sand and leveled (one of the inputs to core design is that it must have one flat face which is the open side of the box). The box is then disassembled very gingerly.

[bananarchy]

Flywheel Rim Segment Casting

I fiddled around with different methods and sequences of disassembling the corebox to best preserve the core - the sand is pretty delicate, and I understand that often additional binders are used in the sand in these applications to aid handling. Something to look into.

The cores are assembled to the mold - one fell to bits on assembly, so I started over and rammed up a second one. Bit of tearout on one exterior corner, but whatever, that surface of the part is getting machined. It was poured (I was actually on one end of the crucible for this one) and the results were.....fantastic! Basically zero issues. There is a tiny bit of flash where the metal got between the mold and the core, but it's extremely thin and very easily snapped off. No sinks, great surface finish. Could not be more pleased.

[bananarchy]

Flywheel Rim Segment Casting

After only some very minor fettling (and significant fondling), the part is looking fantastic - now to repeat the process 11 more times (plus a spare or two). Making the outer surface of the part flat instead of rounded might help make the machining process easier? It will get turned as an assembly of course, but that's a lot of extra material to remove in an interrupted cut. Not convinced future-me is going to be happy with this decision, but time will tell. The next big task is going to be figuring out the machining process - locating the machined geometry relative to the cast features on this part is pretty ticklish. The cast surfaces must line up nicely with the adjacent parts, and there are a lot of ways this can be wrong. Once that's done and datums are established, the rest isn't too bad, but I haven't figured out quite how to go about it yet.

[CI]

Great pattern making/mold making/casting work !
Awesome for sure.

"LinoCure" (tm) by Ask Chemical is a resin binder that is good for use with cores, and/or the entire mold.
Has to be used with very dry sand such as OK85.
Designed for use with iron and steel.
And you can spray on a ceramic mold coat such as "Velacote" (tm), also by Ask Chemical, for a beter surface finish.
Resin-bound cores and molds are very strong, and you can vary the set time by varying the catalyst amount.
Available in 5 gallon quantities.
Sand is not easily reused.
If you have a non-critical flat surface that has little or no draft angle, you can cut a piece of mylar film, and cover the surface with it, which prevents the sand from sticking to the surface.

.

[Jasonb]

The initial castings have come out very well

Can I ask what printer/filament combination you are using and also what surface prep you are doing to the pattern as I can only see the tell tail printer layers on surfaces that won't matter?

[bananarchy]

So slight spoilers, I've actually just printed version 4 of the spoke pattern due to some minor issues with V3. The first three were done on a Flashforge, which I was fighting constantly and which did not have a big enough bed to print the whole spoke in one part. Having to split it up led to inaccuracies and a lot more finishing work. I've been using automotive body filler to smooth out the big stuff, and automotive build primer, with a polyurethane gloss coat on top.

Recently though I pulled the trigger on a Bambu H2S, after having great experiences at work with their X1C, and it's been fantastic. Total game changer. The bigger work area allows it to do the spoke in one go, it's way faster, and the quality is top notch. These were printed on the max quality settings in the basic PLA, and I haven't started the finishing process yet but I expect it to be a whole lot less work than previously.

[Jasonb]

Thanks for the details, latest prints look good.