AECL Research in Manitoba, Canada has developed the INSIGHT steam quality monitor, an instrument used to measure steam quality and mass flowrate. Steam Quality and Mass Flowrate is the energy injected into the ground in an oil recovery project, for example.
The improvement obtained by using the trained network was immediately apparent. Using a conventional linear program, the standard error of estimate (RMS of deviations about the ideal line) for steam quality and mass flowrate are 28% and 0.59 kg/s. Using the trained neural network, the standard error was 8.2% and 0.34kg/s.
A common test set of 26 sets of input data was used and the network was trained on an additional 100 facts.
Later, a similar network was trained and tested all of the INSIGHT monitor calibration data obtained to date (i.e. data from tests at four different facilities collected over a period of seven years using a minimum of six to a maximum of nine different monitors). Here, the standard error of estimate for steam quality and mass flowrate were 7.7% and 0.4kg/s, respectively.
Recently AECL has successfully trained a neural network to return methanol, gasoline and water contents from the RF reflectance spectra of mixtures of these three components. Currently they are investigating the application of a neural network to a-spectroscopy and to the interpretation of on-line chemical sensor signals.
INPUTS