Research on Risk Prediction of Deep Mining in Coal Mines Based on Neural Network and Stacking Integrated Learning Model

• Published in: 2024 IEEE 2nd International Conference on Sensors, Electronics and Computer Engineering (ICSECE) pp. 1353 - 1358
• Main Authors: Du, Shiwei, Yang, Congkai, Jia, Yuqi, Hou, Yixiong
• Format: Conference Proceeding
• Language: English
• Published: IEEE 29.08.2024
• Subjects: Classification algorithms; Clustering algorithms; DBSCAN; Electromagnetic radiation; Ensemble learning; Indexes; Jamming; neural network; Neural networks; Predictive models; Stability analysis; Stacking; Stacking ensemble learning
• DOI: 10.1109/ICSECE61636.2024.10729572

Coal is an important energy and industrial material in China, and monitoring and early warning of rockburst is a key scientific and technological issue in coal mine safety production. Based on the data of electromagnetic radiation and acoustic emission signals, this paper establishes parameter indexes describing the relationship between time span, electromagnetic radiation, and sound wave intensity, and generates characteristic indexes such as data drift and stability. Cluster analysis is carried out by using DBSCAN algorithm, and features are screened by UMBP analytic hierarchy process. Using Topsis algorithm, the jamming signal screening model is established, and the five earliest jamming signal intervals are determined. Using neural network algorithm, the time intervals of five precursor characteristic signals of electromagnetic radiation and acoustic emission signals are determined. On this basis, a classification model based on Stacking ensemble learning is constructed to solve the probability of occurrence of precursor characteristic data in each time period. The analysis shows that the AUC of the integrated model algorithm reaches 86.37%, and the model is stable and effective.