Robust CH-TDLAS sensor based on state-switching adaptive Kalman filtering.

Tunable diode laser absorption spectroscopy (TDLAS) is a highly advantageous technique for trace gas detection, characterized by its low cost, high precision, and compact design. However, in complex field environments, external disturbances such as shocks, vibrations, and temperature fluctuations can induce non-Gaussian abnormal noise in the system, significantly affecting measurement accuracy. In this paper, a robust gas concentration measurement method based on state-switching adaptive Kalman filtering is proposed. First, the correntropy criterion is introduced into the adaptive Kalman filter based on a forgetting factor to suppress abnormal noise. Subsequently, an adaptive state switching mechanism based on chi-square detection is established. The stability of the system is identified by the discretization of measurement data within a fixed-length sliding window, and the size of the width of the correlation entropy core is adaptively adjusted, enhancing the robustness of the system in the field environment while ensuring the real-time performance of state tracking. A TDLAS sensor based on state-switching adaptive Kalman filtering was established for methane (CH) concentration detection. In practical measurement experiments under complex scenarios, involving dynamic response, abrupt noise interference, and mismatch noise interference, the root mean square error (RMSE) of gas concentration of the proposed method demonstrated improvements of 61.6%, 25.9%, and 18.5%, respectively, compared to traditional methods.