A novel method for resisting missing monitoring data of mining subsidence by integrating EMD and RBF.

Due to the lack of field monitoring points, the monitoring data is incomplete, making it difficult to obtain more complete characteristics of mining subsidence. To address this issue, a novel method for interpolating missing subsidence points is proposed. Firstly, the subsidence data are decomposed into multi-frequency domains utilizing the empirical mode decomposition (EMD) method. Subsequently, the missing subsidence data are reconstructed by combining the radial basis function (RBF) with phase space reconstruction technology. The interpolation accuracy is then analyzed under various conditions, including different mining degrees, sample numbers, and observation point locations. The results indicate that the interpolation accuracy increases with the mining degree coefficient. During the active stage of surface subsidence, a larger sample number correlates positively with higher interpolation accuracy. However, in the recession stage, the impact of sample number on interpolation accuracy is less pronounced compared to the timeliness of sample acquisition. The location of observation points has minimal effect on interpolation accuracy. By applying this method to the 833 working face in the Shanjiacun coal mine, the mean absolute percentage error (MAPE) is less than 20%, and the mean squared error (MSE) is below 0.5 mm, objectively demonstrating the method's effectiveness and engineering applicability. These research findings offer valuable technical support for mine safety production and ecological environment protection. [ABSTRACT FROM AUTHOR]