金属矿深部开采岩爆危险预测的GA-ELM模型研究

为对金属矿山深部开采时岩爆的危险性进行预测,在总结深部开采岩爆发生机理的基础上,综合选取影响岩爆发生的3个重要因素作为岩爆预测的判别因子。搜集国内外金属矿深部开采岩爆的实例作为训练样本,引入极限学习机算法(ELM),针对该算法的不足,采用遗传算法(GA)对其相关参数进行优化,建立了岩爆预测的GA-ELM模型,并与单一ELM模型进行对比。利用该岩爆预测模型对一典型金属矿深部开采进行岩爆预测,结果与实际情况相吻合。研究结果表明,岩爆预测的GA-ELM模型训练效果及泛化能力均优于单一ELM模型、SVM模型及传统的BP模型,且该模型能够对金属矿深部开采的岩爆进行准确有效地预测,具有一定的工程应用价值。

Rockburst Prediction with GA-ELM Model for Deep Mining of Metal Mines

In order to predict risk grade for rockburst during the deep mining of metal mines, three major factors to influence rockburst were selected as the discrimination factors for rockburst prediction based on the mechanism of rockburst in deep mining. Engineering practices of rockburst in deep mining at home and abroad were collected as training samples. Considering the disadvantage of the extreme learning machine(ELM), genetic algorithms(GA) was used to optimize the related parameters of ELM and a GA-ELM model for rockburst prediction was established, which was also compared to the ELM model. It was found that the prediction with GA-ELM model for the typical deep mining of a metal mine showed the result was consistent with the actual situation. Research indicated that, GA-ELM model was superior to the single ELM model, SVM model and the BP model in terms of training effect and generalization capability. Moreover, GA-ELM model can provide an effective and accurate rockburst prediction for deep mining of metal mines, which can be applied into practical engineering project.