err yep... what Tom said.
In fact it is very simple if you have the parts available, or are prepared to do a bit of rewinding.
1. The transformers need to be stripped, the stacks glued together, and rewound as a single transformer.... you should have three... or at least two.. never know with these folks.... three would be nice. Use 1/3 the turns for the 240v/110v winding as in original, and twice the turns on a single original transformer ( 24v turns) for the 48v winding... thats step 1... simple but time consuming.... if you only got two transformers, then 1/2 turns for the HV winding, and the same total turns for the 24v winding with both cores .
Eg. 240v unit... if you had 5 turns: 240 turns on a single original, and we have three, they will be series primary, and parallel secondary... so it will look like 15 turns of primary to 240 turns of secondary. When we stack three, we divide the turns by 3... so 5 turns to 80 turns for 24v but we want 48v so multiply the primary by two, and we see 10 turns to 80 turns on the three stack transformer for the 48v version..... the HV winding needs to be wound with three in hand for the same power handling as the original three ( more power actually, as we have effectively gotten rid of a lot of copper, almost 1 entire transformer worth) in short, you want to end up with about an 8:1 ratio for 48v 240v transformer.
Plan B if you don't want to unwind the long HV windings... and rewind three in hand, is to just stack the three transformers together after removing the big LV windings... strap them together as a single cylinder, and get as big a welding or battery cable as you can fit in there to wind twice the turns of a single Lv winding off the original. This will give you a better transformer than the originals, but not as good the first conversion. The top conversion will yield a significantly better transformer than the originals.
2. Your main fet board..... depends on what fets they used. If they used IRFB4110, then no change, if they used fets with less than 75v S/Drain then you need to frreplace them with 24 IRFB4110... probably $24.
3. The control card... either buy a new 48v one off ebay seller ( get the 15kw 48v 110 or 240v one) and install... now your finished.
OR you will have to track down the voltage sensing for the main chip, and shunt that, so that it sees half the voltage it would currently see.... and possibly change the capacitor next to the power supply with a 100v rather than a 50v one.. although I think they are all 100v now.... they used to use 50v in some of the 48v ones.. and they explode quite well...
So it is easy of you say it quickly.... or maybe the wild cat is a better option if you don't like playing with transformers.
...............oztules
Edit: It is easy to say get 24 x 4110's, but in practice it is more difficult to do this for a reasonable outlay.
If you buy off ebay or aliexpress , it is normal to be fleeced.. ie I bought 300 4110's hoping they wouldn't be, but knowing they probably would be fakes.... but if you test the Rds on of each of them, you normally end up with a selection of pretty good fets for very little money... at least 200 of those tested less than .007ohms, which for 100v fets is pretty good... not as good as the advertised .0037R, but good none the less.....
So Test them and keep the like with the like, and use the lowest rds on the low side switchers, and the higher rds on for the high side switches..... some of the ones I got were terrible at .024R... effectively a BUK455 60A... but at 100v... so even though they were "lousy" they are better than my stock BUK's.. so it is not all sour grapes.
It looks like they are the rejects from the batches, and some cunning devil sells them for next to nothing ... and they are fakes.... but mostly good enough for most projects. The voltage seems to be reliable, so it looks like the insulated gates are not quite right, (thickness and surface area) and this effects both the capacitance and the Rdson.... but the die types seems to keep the voltage ok........ or maybe they know most won't notice running slightly warmer, but would notice right away if they blew up on switch on.
Interestingly, the lower Rdson ones have lower gate capacitance, so turn on easier.... and this may actually help the heating stakes more than the rds on being wrong... I don't know, it will depend on the current in your driver.
Use a constant current source, and drive the gate at at least 11v, and measure the millivolt drop across the d-s pins right at the case... do the math, and establish the new rds on... then mark it on the case!... don't just create piles of the same sizes... they will get messed up somehow, and you have to do it all again... wonder how I know....