John - your problem isn't easy. If you don't know the characteristics of the waveforms you may encounter you need a true RMS converter.
Math is great on known waveforms - you can implement an algorithm based on a Fourier series representation of any KNOWN waveform to arrive at its true RMS value - but if the waveform is unknown or impure, forget it.
By definition, the heating effect (ie the temperature rise of a resistance) of a current is directly proportional to its RMS value. So, by passing a current that represents the waveform of interest thru an appropriate resistance - and measuring its temperature rise, you will achieve an accurate measure of the RMS value of the current.
Usually, it's too tough for people like us to design these sort of schemes, so I'd suggest you look for a TRUE RMS converter from people like Burr-Brown etc.
Let me know how you got on
good luck & regards
Allan Coyne