Feature extraction for safety coalmine gas concentration prewarning based on HHT and SVD

The gas explosion risk caused by gas overrun is a serious threat to safety production in coalmines in China, quantitative determination of the production impact of the gas concentration is very important for reliable safety prewarning in coalmines. In view of the mine gas concentration varies with the production process. The data sequence of real-time gas monitoring were decomposed into the Intrinsic Mode Function (IMF) with different time scales by using Empirical Mode Decomposition (EMD), then the noise signals in it is eliminated by using Hilbert spectrum analysis, furthermore, combined EMD with Singular Value Decomposition (SVD), the singular value sequence is extracted to reflect the intrinsic feature of the gas concentration. According to the singular value extracted by work shifts, the maximum feature value and the mean increment of the gas concentration affected by production process are determined; according to the singular value extracted by hour, the time of the gas concentration decreasing to a stable level after production stop is determined, and then the prewarning level is determined. The case study shows that the mean variation trend of the gas concentration obtained by feature extraction conforms to the theoretical characteristics of mine gas flow in ventilation roadway. The prewarning parameters obtained by feature extraction are quantitative, and their values can change dynamically with the production conditions, and the determination of prewarning threshold is more refined. the mean reduction of the gas concentration between production shift and non-production shift, and the time of the gas concentration decreasing to a stable level after production stop are important basis for safety prewarning decision under the normal mine ventilation conditions.