Dynamic Monitoring and an Early Warning Model of a Floor Water Disaster: A Case Study

The water hazard of the coal seam floor is a major threat to safe coal production in China. To improve the accuracy of water hazard predictions, a water inrush risk predictive model was constructed using PSO-SVM. Historical monitoring data were added to the basic database in a timely manner to narrow the difference between the monitoring data and predicted results. The optimized database was used for neural network model training. The prediction model was improved by establishing a database self-optimization and model self-learning process (SOMSP). The PSO-SVM model and the SOMSP was used to predict the inrush risk for 23 groups of floor water inrush cases from the north China mine area. The initial accuracy of the model was only 25% for the first 19 data groups, which were used as the basic training data to predict data groups 20–23. Using the SOMSP, the accuracy of the water inrush risk of the coal seam floor was increased to 100% (3/3). Thus, the accuracy of the predictions was greatly improved by the SOMSP.