TEMPERATURE ANALYSIS OF COHERENT ANTI-STOKES-RAMAN SPECTRA USING A NEURAL-NETWORK APPROACH

A neural network trained with clustered data has been applied to the e xtraction of temperature from vibrational Coherent Anti-Stokes Raman ( CARS) spectra of nitrogen. CARS is a non-intrusive thermometry techniq ue applied in practical combustors in industry. The advantages of clus tering of training data over training with unprocessed calculated spec tra is described. The method is applied to CARS data from an isotherma l furnace and a liquid kerosene fuelled aero-engine combustor sector r ip. Resulting temperatures have been compared with values extracted fr om the data using conventional least squares fitting and where possibl e, mean temperatures measured by pyrometer and blackbody The main adva ntage of the neural network method is speed, with the potential for on line temperature extraction at the spectral acquisition rate of 10 Hz using standard PC hardware.