Research Of Earthquake Disater Warning In Mine Drainage Based On Extreme Machine Learning

The disaster monitoring and early warning of mine dumps is still in its infancy,but the frequent occurrence of mine accidents in mine dumps has caused a large number of casualties,property losses and adverse social impacts.With the various types of sensors applied to the safety monitoring of mine dumps,the short-term early warning mechanism for the safety situation of mine dumps is realized,which has important research significance for improving the initiative of relevant enterprises in dealing with corresponding disasters..In this paper,the safety warning status of mine dumps is researched,and the mine dumping monitoring and early warning system of B/S and C/S hybrid structure is developed.The real-time data of each characteristic index of mine dump is realized.Real-time monitoring needs of various indicators of the dump.Firstly,through the principal component and correlation analysis of the collected historical data,the early warning targets of mine dumps with surface displacement and multivariate fusion are established.Secondly,the data of mine dumps are analyzed and excavated.The integrated learning prediction model based on AdaBoost particle swarm optimization extreme learning machine is established.The optimization of particle swarm optimization algorithm avoids the limitation of artificial selection of input parameters of extreme learning machine and the integration idea of AdaBoost to improve the prediction accuracy of extreme learning machine.Experiments show that the model has achieved good results in the off-line prediction stage.Thirdly,considering that there is a large amount of accumulation and dynamic fluctuations in the relevant data of mine dumps in the long-term,the offline model can not meet the real-time forecasting requirements.Self-encoding adaptive online cyclic limit learning machine model consisting of two layers of self-encoding network and one layer of RNN network.Through the addition of RNN network,self-encoder and adaptive forgetting factor,the relevant weight of the model can be New input data stream is determined,while eliminating interference and increase of noise data The anti-interference of the strong model shows that the proposed model performs better than other related models in terms of prediction accuracy and efficiency.Finally,in order to cope with the future accumulation of mine dump data,a distributed adaptive online limit learning method is proposed.Machine algorithm model,predictive tasks through distributed systems,experiments show that the effectiveness of parallel algorithms is improved,and the efficiency of distributed systems in massive data processing.