Suppression of cavity mode noise in off-axis ICOS based on a Kalman filter optimized by NRBO-SVMD and LSTM

Residual cavity mode noise represents a major constraint on the detection accuracy of trace gases in off-axis Integrated cavity output spectroscopy (OA-ICOS). The Kalman filter (KF) is employed for the noise suppression owing to its strong real-time dynamic estimation capability. However, its performance is limited by the quality of the observation signal and the accuracy of the prediction signal in OA-ICOS. To overcome this limitation, this paper proposes a cascaded denoising architecture designed to enhance both signals: firstly, to improve the quality of the observation signal, the Newton-Raphson-based optimizer (NRBO) is employed to adaptively determine the optimal value for the parameter in sequential variational mode decomposition (SVMD). SVMD then decomposes the original signal to preliminarily separate noise from the useful component. Subsequently, to enhance the accuracy of the prediction signal, a long short-term memory (LSTM) network is utilized to construct a prediction model. Finally, the Kalman filter dynamically integrates these optimized observation and prediction signals through an adaptive weighting mechanism, achieving optimal state estimation. The application of the NRBO-SVMD-LSTM-KF algorithm significantly improved signal quality: the signal-to-noise ratio (SNR) increased from 21.159 to 33.715, the standard deviation (STD) decreased from 9.89576 to 0.99406, and the peak-to-peak value was reduced to 6 ppb, accompanied by a 91.43% suppression of signal fluctuations and markedly improved system stability. Compared with conventional filtering algorithms and hybrid methods, the proposed algorithm demonstrates superior noise suppression capability, offering an effective solution for suppressing residual cavity mode noise in OA-ICOS systems.