Near-Infrared CO Sensor Based on Small Multi-Pass Gas Cell and GCRA-BiLSTM Model in Underground Parking

Abstract

Due to the influence of automobile exhaust in underground parking lots all year round, a large amount of carbon monoxide (CO) gas accumulates in underground parking lots, endangering the safety of human lives. Based on tunable diode laser absorption spectroscopy (TDLAS) technology, a CO gas detection system is designed to be applied in underground parking lots. A distributed feedback baser (DFB) laser was used to scan the CO absorption peak at 1.566 μm, avoiding interference from other background gases. A small multi-pass gas cell (MPGC) was designed with a volume of only 704 cm³ and an effective optical range of up to 26 m. An improved wavelet thresholding denoising-complete esemble empirical mode decomposition with adaptive noise (IWTD-CEEMDAN) denoising algorithm was designed to filter out the effects of background noise, thereby improving the signal-to-noise ratio (SNR) and lowering the detection limit of the system. A greater cane rat algorithm-bidirectional long short-term memory (GCRA-BiLSTM) concentration inversion algorithm was designed to improve the accuracy of the system concentration inversion. According to the Allan variance, the limit of detection (LoD) of the sensor is 11.57 ppm at 1 s and 0.124 ppm at 117 s. A series of experiments are carried out on the sensor, and the experimental results show that the sensor has good linearity, stability, and LoD, which can realize the real-time monitoring of underground garage CO, and it has an important application value.