Ultra-Low Sampled and High Precision TDLAS Thermometry Via Artificial Neural Network

Water vapor temperatures in a heating device were measured by ultra-low sampled and high-precision tunable diode laser spectroscopy (TDLAS) Bichromatic distributed feedback (DFB) lasers were used as light sources. Sampled data at an ultra-low rate was collected after laser passing through a low-pass filter and served as the inputs of an artificial neural network. Classical direct absorption spectroscopy using the line-shape fitting method provided the training dataset, i.e., the integrated absorbances. The proposed method required an ultra-low sampling rate, i.e., only 1/50 of the classical method, but its calculation speed was nearly 14,000 times faster. Also, the proposed method yielded satisfying estimates at temperatures uncovered by the training dataset and was insensitive to random noises.