Thread: Then and now
View Single Post
  #196   Report Post  
Posted to alt.home.repair
dpb dpb is offline
external usenet poster
 
Posts: 12,595
Default Then and now

The Daring Dufas wrote:
....

In other words your sensors were passive listening devices not active
like the ultrasonic flow sensors I'm familiar with?


Yes, as I said they were/are simply high frequency (75 kHz resonance)
commercially available accelerometers...

... I'm guessing you
were looking for a semblance of a pattern in the white noise of the
chaos. Am I getting warmer? :-)


Except white noise implies a stochastic process whereas a chaotic
process, while not strictly repeatable, is not stochastic. Hence,
typical statistical measures used for such processes are not effective
and the measures used in the analysis are, as mentioned, nonlinear and
then combined in other nonlinear ways for prediction. (I know, that's a
lot of mumbo-jumbo unless one already knows the answer, but as I say,
the specifics are proprietary so can't really reveal much more about the
details).

Suffice it to say that there is information buried in the audible and
supersonic noise of the flow and that can be related to the actual air
and coal flow rates in a given pipe over a range of operating conditions
and air:fuel ratios and flow rates by a set of specific operations on
the recorded waveform. One important feature of these computations is
that they all produce quantities that are independent of the actual
magnitude of the signal itself (iow, they're self-normalizing). This is
a key feature in that it means that simply a level change from a
location difference doesn't affect a given signature. It also means, of
course, that simple measures such as the mean aren't what is giving the
actual correlations. But, from those correlations a prediction of
flow is possible for a given new set of measures computed for the same
pipe from any set of operating conditions and this has been shown to be
valid over a range of operating conditions and at various power plants
of differing sizes and styles and manufacturer (albeit the correlations
are at least to this point plant-specific, the measures used in those
are for the most part the same ones of of the total set of those
identified as candidates).

--