Double filter noise reduction algorithm optimized by RBF neural network for laser absorption spectroscopy

Abstract

Acetylene(C₂H₂) gas sensors are vital for monitoring the safety of atmospheric as well as industrial environments. In this paper, a high-sensitivity trace gas detection sensor for C₂H₂ has been developed using tunable diode laser absorption spectroscopy (TDLAS). As we known, the accuracy and sensitivity of TDLAS is largely limited by data noise as well as concentration inversion algorithms as a calibration-free gas detection technology. In order to solve this problem, a dual-filter noise reduction algorithm based on neural network optimization was proposed, which is applied in the process of spectral data processing and concentration inversion. The algorithm is initially tested by using simulated absorption spectra and compared with traditional single filter signal processing algorithms. Subsequently, experimental tests were carried out using standard-concentration gases as well as high-precision long-path gas absorption cell. The experimental results demonstrate that the proposed algorithm increases the signal-to-noise ratio (SNR) from 14.31 dB to 22.07 dB, a 7.76 dB improvement. The standard deviation of the inversion concentration is lowered from 12.53 ppm to 3.25 ppm, representing a 3.86 times improvement. Furthermore, Allan deviation results show that the data filtered according to the proposed technique has a detection sensitivity of 0.83 ppm at the optimal integration time of 19 s. These results demonstrate that the data processed by the algorithm can more precisely predict concentration values, providing significant support for increasing the performance of gas sensor systems.