As a 'last resort' method, you could apply a current limited power supply to ground, and the PCB point that is shorted to ground - gradually turn the current up, and look for what smokes or gets hot. If you're lucky a component will get hot, and that's likely to be what's at fault - if you're unlucky the PCB itself will get hot, and it's likely an internal short on a multi-layer board.
But before you get to the 'last resort' try measuring the actual resistance of the short with a low-ohm meter, and trace along the track till you find the lowest resistance point. With this technique you can tell which end of a track is closest to the short - and also test directly from a component pin, if it's a transistor, and the shorted pin read 50milliohms, and the other pins read higher than that, then obviously it's not the transistor at fault. You can also use the same technique to actually remove components, or lift connections.
I purchased a low-ohm meter specifically for that use - and actually had occasion to use it a few weeks ago. It was on a board I'd designed, and had just populated one that didn't work properly - and a few tests showed the MAX7219 wasn't working correctly (two of the outputs were shorted together - and yes, I tried changing the chip before going further). So I dug the low-ohm meter out, measured along the tracks and was able to locate the area where the short was - and using a magnifier headset I was able to see a hairline of copper between tracks that hadn't etched away. A quick cut with a sharp knife, job done
But without circuits, and full service manuals, it's usually a pretty difficult job