Crazy idea, maybe you can make your own induction heating crucible and mix your own alloys, Applied Science style.
I mean, there's a few steps missing there, but yes, generally possible...
I've actually wanted to do that for quite a long time. I was doing amateur foundry in high school, and that's what gave me the idea to make an induction heater. Alas, I need something other than this apartment to do that in..!
Add to that: grinder, rolling mill, probably a regular mill, annealing furnace (that's at least trivial if I've got the first part down
), shear, and anything else to make finished parts (brake? dies? brazed or spot welded items?).
I wonder how tricky it is to get a magnetically soft alloy. The amount of carbon and other alloys should be very small, AFAIK. With such little carbon there's also a very fine line before picking up oxygen instead. Which dissolves quite well into molten iron, eventually separating as molten FeO floating on top. Cool fact, all iron oxides melt slightly below the pure metal. Which looks really interesting when you've got a piece of steel at that temperature in a neutral to oxidizing atmosphere: it starts dripping!
So, inert gas purging may be desirable; add welder and gasses to the list.
But really, that's at or under $10k of equipment, to do small scale production of real materials, not a huge cost as capital goes.
Composition is easier to control, I think, with a large heat. The big guys can also take samples in the middle of a run, go over and XRF it, and adjust it as needed. Which, even adding XRF capability isn't a huuuuge investment as they show up relatively cheap from time to time. But I'd really be wanting to think about finding customers for specialty alloys at that point, and spinning off the business...
Edit: With the recent advances in power semiconductors and capacitors, it should be cheaper and easier than ever to make a decent induction heating crucible/forge. Makes me want to make one now that I think about it... it's just a resonant locked LC converter after all...
It's really not so hard, a PLL gets resonance tracking by default, the output of which can be pulled to control current, voltage, power, whatever. Just add on those sensors (typically a voltage divider, current transformer, and active rectifier and filter to get the envelope of them), and error amps, and you've got it. Don't forget fault circuitry, a peak current detect is worthwhile, and desat protection for the inverter can save literal buckets of transistors.
Compensation is rather difficult in this method, as the complex pole in the loop response corresponds to the difference between driven and resonant frequencies. Consider taking the envelope of this and trying to control it:
Stimulus is FSK (open loop, fixed frequency drive), with the lower amplitude being higher above resonance. Evidently, 3.3kHz above, and the higher amplitude, close to 1kHz. Resonance looks to be about 21kHz then.
The alternative is always locking to resonance, and modulating the supply -- so you need two converters. Still easy enough, just more stuff to do, and a bit less efficiency.
Note that you can even do real-time impedance measurement, given a voltage and current sensor. You can solve for the network components in real time, and infer properties of the work; maybe even temperature or distance or cross section. (Or with more refinement: eddy current testing is used to detect variations or defects in metals.)
If you're interested enough to do experiments, I recommend a series resonant load. The resistance is usually quite low, making it easy to match with a one-turn toroidal transformer,
or maybe two. (Note that there are better ways, with lower leakage inductance, but this is pretty easy to construct.) Also, as a simple series circuit with a known component (capacitor), you can solve for L and R in real time, if you like.
But I digress; given a load with fixed Tc, and modest coupling, control may not even be needed. I think this could be useful for certain dedicated, well-defined loads -- certainly, Metcal found one such case. I suppose for Tc nearly room temperature, it could be used for just keeping things warm, and up from there, anything like: preheating; cooking, frying or baking food; soldering, etc. If such materials exist, it would be a boon for development of applications in these spaces.
Tim