摘要
为了更加准确地预测硫化矿自燃安全性,综合考虑硫化矿自燃倾向性及火灾后果严重性,将硫化矿自燃安全性划分为9个等级,并选取矿山含硫量、矿山含碳量、矿石温度、矿石堆放时间、采场人员数量、氧气浓度和采场矿层厚度作为评价因素集。利用主成分分析法(Principal Component Analysis,PCA)对94个采场样本数据进行降维处理,得到包含70%以上原始信息的3个主成分。将降维后的84组数据作为基于径向基函数神经网络(Radial Basis Function Neural Network,RBF)预测模型的训练样本,10组数据作为检验样本进行硫化矿自燃安全性预测。最后分别利用十折交叉验证法和留一法对94组检验样本的自燃安全性预测结果进行检验,得到硫化矿自燃安全性预测准确率分别为92.55%和91.49%。研究结果表明:PCARBF网络模型对硫化矿自燃安全性的预测性能良好,且优于未经主成分分析的结果。
Spontaneous combustion of sulfide ore will cause a series of environmental,safety,and property hazards.It is of great practical significance to predict the tendency of spontaneous combustion of sulfide ore and the severity of fire consequences more accurately for the realization of more safe and efficient mining of sulfide ore. In this paper,seven factors affecting the spontaneous combustion tendency of sulfide ore were comprehensively considered as the evaluation index factors,including mine sulfur content,mine carbon content,ore temperature,ore stacking time,the number of stope personnel,oxygen concentration,and stope ore layer thickness.The spontaneous combustion safety of sulfide ore was divided into nine grades,representing different spontaneous combustion tendencies and severity of fire consequences. 94 sets of actual stope data were collected,and the principal component analysis(PCA)was used to reduce the dimension of the 94 sets of stope data.Three principal components containing more than 70% of the original information were obtained.84 sets of data after dimension reduction were used as training samples of the radial basis function neural network(RBF)prediction model,and 10 groups of test samples were used to establish the PCA-RBF self-ignition prediction model of sulfide ore. The 10-fold cross-validation method and leave one-out method were used to verify the prediction results of the PCA-RBF model with the actual results. The prediction accuracy of the PCA-RBF model is 92.55%,and the correlation coefficient is 0.94. The prediction accuracy of the PCA-RBF model is91.49%,and the correlation coefficient is 0.97. Both of the two verification methods show that the PCA-RBF model has good applicability to the prediction of spontaneous combustion safety of sulfide ores.The results of a small amount of prediction deviation are also less different from the actual results,and the overall prediction accuracy is higher than that of the RBF model. The results show that the radial basis function neural network based on principal component analysis has good prediction performance for the spontaneous combustion safety of sulfide ore.The prediction accuracy of the sample is above 90%,and the correlation coefficient is greater than0.9,which is better than the results without principal component analysis.PCA-RBF model can be used to predict the grade of spontaneous combustion safety of sulfide ore,which can guide the safety production of mine.
作者
杨珊
李文文
陈建宏
YANG Shan;LI Wenwen;CHEN Jianhong(School of Resources and Safety Engineering,Central South University,Changsha 410083,Hunan,China)
出处
《黄金科学技术》
CSCD
2022年第6期958-967,共10页
Gold Science and Technology
基金
国家自然科学基金青年基金项目“基于人工智能的矿山技术经济指标动态优化”(编号:51404305)资助。
关键词
硫化矿
自燃倾向性
火灾后果
主成分分析
RBF神经网络
等级预测
sulfide ore
spontaneous combustion tendency
fire result
principal component analysis
radial basis function neural network(RBF)
grade prediction