I assume there's an associated warning light or something electrical you can tap-into? Or, assuming that each of these warning systems has it's own-separate buzzer/sounder, you can tap-into that signal.
I wouldn't trust sound detection. It's not as easy as you might think when you've got noise/other sounds or when the loudness can vary. Just as an example, you can listen to a song and you can hear the singer and the guitar and you can tell the difference. A computer can't do that... A computer can't remove the guitar from the recording without "damaging" or removing the vocals and other instruments.*
The audio alarms are for humans. What is the Arduino going to do with the information?
I've never used an LM567, but it looks like you'd need one for each tone/frequency. You can also build filters with op-amps or use [u]FFT[/u]/FHT in software.
- There are "vocal removal" techniques, but these mostly rely on left/right/center location in a stereo recording and they don't work that well.