Signal denoising and pickup are crucial for extracting information from collected data. We developed an integrated method using deep neural network (DNN) to solve these two procedures simultaneously. The DNN is designed by two homo-structured encoder-decoder networks in series to facilitate the accuracy and efficiency of signal processing. Results show that the method has an optimal performance on dealing with microseismic signals that contain various types and intensities of noise, even the signals and noises share the same frequency band. The results of signal pickup are more in line with the actual duration of microseismic signals. Compared with existing methods, this method removes the noise with a minimum level of waveform distortion. It is ideal for recovering the micro-seismic information while maintaining a good capacity for pickup when the signal-to-noise ratio is low. The method has great potential to be extended to the study of exploration seismology and earthquakes.

Time series prediction involves learning existing observation data of a parameter and predicting its future evolution, with applications in many fields. Based on a machine/deep learning method, it is useful to predict the evolution trend of microseismic parameters during rockburst development. Our study explores key microseismic indices that help describe the development process of rockbursts based on abundant rockburst data obtained from deep underground engineering activities. On this basis, the structure of a deep convolution neural network is modified, and its input and output modules are improved, to realize a univariate, multivariate input and a single-step, multi-step output. Finally, various new models of the microseismic multi-parameters for time series prediction are proposed, including a univariate prediction model, a multiple parallel series model, a multiple input series model, and a multivariate multi-step prediction model. Model training, testing, and interpretation of the rockburst risk and a comparative analysis of the different models are carried out for the complete process of multiple rockburst disasters. The results show that the proposed models can predict the evolution trends of various key characteristics during a rockburst and ensure the timeliness and accuracy of rockburst risk prediction. They provide a new research idea for the prediction and early warning of rockburst disasters in deep underground and mining engineering.