Chemistry 101.
Heat up any element and you force the electrons into a higher "orbit" or energy state.  Eventually, that electron will
fall back to its natural state, emitting a photon of a characteristic wavelength in the process.  White light is a
combination of all the available wavelengths combined.  Split out all the wavelength and you get a rainbow.  If the
light comes from burning a particular set of elements, that rainbow will exhibit "bands" or bright stripes of one
particular color (ie, wavelength).  Every element on the periodic table can be identified by a particular set of bands
that are more accurate than finger prints.  Astronomers use this phenomena to accurately identify what elements are
burning in a distant start.

Carbon has a distinctive fingerprint that is very different from the iron, nickel and chromium in a glowing exhaust.
See page 3 of

http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=7&ved=0CGAQFjAG&url=http%3A%2F%2Fwww.tempe.mi.cnr.it%2Fzizak%2Ftutorial%2Fcairol06-flame-emission.pdf&ei=t-QbUMenOczWiAKR6oGYAg&usg=AFQjCNHah9qiwVWEvkBeVo1YXXaTxnpWgA&sig2=_nZhVPA8PD3CiUNUe4Rn7A

To build a reliable flame detector, you'd build a sensor that would respond to wavelengths of a particular band that is
characteristic of carbon, and then add an IR detector that just confirms there is a lot of heat present.  The article
even suggests that it would be possible to tell the difference between an atmospheric gas flame and a welding flame,
because of the nitrogen in the air.

Dave wrote:
>
> I would say that the glowing red headers wouldn't set EITHER off. The
> NEWER ones are less likely to be fooled by a welders arc.. remember,
> these devices are meant to be used in refineries as fire detection
> triggers to deluge sprinkler systems..... welding and other hot work
> happens in that setting.
>
> I remember at a fly in at Pecan Grove about 8 years ago or so, someone
> was presenting about the detector and we were ALL surprised when in the
> middle of the presentation the detector "detected".... and we all
> scratched our heads until I noticed a propane radiant heater with a
> sunflower style burner in the corner 40 feet away was our source.
>
> The detectors use UV spectrum for flame/combustion detection. They use
> IR for crosscheck/validation, not primary detection. Thats my laymans
> understanding of it.
>
> Dave
>