Objective In the process of exploiting combustible ice, the temperature in the production string can decrease due to the influence of depressurization and hydrate decomposition. This can produce a temperature and pressure environment conducive to the secondary formation of methane or carbon dioxide hydrate, which results in the risk of string blockage. Therefore, it is very important to monitor the gas concentration in real time during drilling in order to determine from the gas concentration whether a secondary hydrate is starting to form and to select measures flexibly to prevent this from blocking the pipe string. However, due to the sizes and depths of the wells, it is difficult to complete downhole work for long-optical-path gas-absorption pools, such as a White pool or a Herriot pool. For an optical path with limited effective absorption, it is important to ensure the smooth progress of combustible ice mining by using digital-filtering methods to improve the signal-to-noise ratio of the system and reduce the minimum concentration necessary for environmental gas monitoring. Traditional denoising methods, such as the Kalman filter, wavelet transform, and empirical mode decomposition, are of limited utility because of the problems of mode aliasing and endpoint effects. In the present study, we have combined a variational mode-decomposition (VMD) algorithm with the Savitsky-Gorai (S-G) filtering algorithm to produce a VMD-based filtering algorithm that effectively solves the modal-aliasing problem and removes system noise. We expect that our basic strategy and findings will be useful for the constant monitoring of gas concentrations during drilling in combustible ice mining.
Methods This paper proposes a VMD-based filtering algorithm based on the VMD algorithm and the S-G filtering. First, it is necessary to determine the number of decomposed modes. We choose different numbers of modes for decomposing the gas signal using VMD. We then apply the fast Fourier transform (FFT) to the decomposed modes, and we determine the optimal number of decomposed modes according to the amount of mode overlap. By calculating the Pearce correlation coefficient between each decomposition mode and the original signal, it is possible to judge whether the effective signal is a low-frequency mode. After decomposition a low-frequency mode is selected for S-G filtering, and the residual high-frequency noise after VMD filtering and the low-frequency system noise are filtered out. In addition, we calculate the system signal-to-noise ratio, linear correlation coefficient, and minimum detection volume fraction to evaluate the performance of the algorithm.
Results and Discussions In our simulation experiment, the absorption line of CH4 at 1653.72 nm is taken as an example, and the pure signal and the dye-noise signal are obtained (Fig. 1). The dye-noise signal is processed using the VMD-based filtering algorithm to yield the filtered signal [Fig. 4(b)]. The noise in the contaminated signal is thus effectively filtered out, and the signal-to-noise ratio increases from 7 dB to 20.1 dB. Then we applied the VMD-based filtering algorithm to the actual gas-monitoring instrument used in drilling to mine combustible ice in order to obtain the gas signal before and after noise reduction (Fig. 9). The signal-to-noise ratio of the gas signal increases from 8.5 dB to 21.7 dB. Gases with volume fraction ranging from 200×10 -6 to 500×10 -6 at intervals of 50×10 -6 were selected for detection in order to determine the relationship between the second-harmonic amplitude and the volume fraction (Fig. 10). The linear correlation coefficient of the instrument is increased from 0.9633 to 0.9940 by using the VMD-based filtering algorithm, and the minimum detectable volume fraction decreases from 85.2×10 -6 to 6.7×10 -6.
Conclusions In this study, we have proposed a VMD-based filtering algorithm to improve the signal-to-noise ratio for gas detection by combining the VMD algorithm with the S-G filtering algorithm. The problem of modal aliasing and end effects in previous gas-noise reduction algorithms is well solved with this approach, and the effective signal is separated from the noise signal. We applied the VMD-based filtering algorithm to the actual gas-monitoring instrument used during drilling to mine combustible ice, and the experimental results show the new algorithm produces substantial improvement. Our research has thus shown that this VMD-based filtering algorithm effectively improves the signal-to-noise ratio and decreases the minimum gas volume fraction detectable with the instrument, making this method an effective tool for the field of spectral signal processing and gas-concentration monitoring.
