Borehole extraction is the basic method used for control of gases in coal mines. The quality of borehole sealing determines the effectiveness of gas extraction, and many influential factors result in different types of borehole leaks. To accurately identify the types of leaks from boreholes, characteristic parameters, such as gas concentration, flow rate and negative pressure, were selected, and new indexes were established to identify leaks. A model based on an improved naive Bayes framework was constructed for the first time in this study, and it was applied to analyse and identify boreholes in the 229 working face of the Xiashijie Coal Mine. Eight features related to single hole sealing sections were taken as parameters, and 144 training samples from 18 groups of real-time monitoring time series data and 96 test samples from 12 groups were selected to verify the accuracy and speed of the model. The results showed that the model eliminated strong correlations between the original characteristic parameters, and it successfully identified the leakage conditions and categories of 12 boreholes. The identification rate of the new model was 98.9%, and its response time was 0.0020 s. Compared with the single naive Bayes algorithm model, the identification rate was 31.8% better, and performance was 55% faster. The model developed in this study fills a gap in the use of algorithms to identify types of leaks in boreholes, provides a theoretical basis and accurate guidance for the evaluation of the quality of the sealing of boreholes and borehole repairs, and supports the improved use of boreholes to extract gases from coal mines.
© 2022. The Author(s).