基于BP神经网络和SVR的Fund■o尾矿坝排水数据预测对比研究
|
摘要
针对尾矿库运行过程中安全预警问题,选取2015年巴西Samarco铁矿溃坝事故案例,研究BP神经网络和SVR方法在排水数据预测的适用性。综合分析了排水数据的复杂且非线性的特点,以库水位、降雨量和干滩长度为输入特征,采用上述2个模型对尾矿坝排水数据进行预测。研究结果表明:基于BP神经网络预测结果的最大相对误差不高于4.35%;基于SVR算法的最大相对误差不高于9.21%;Fund■o坝的排水预测结果是可行的,BP神经网络的预测精度更高,而SVR模型的运算速度更快。研究结果可为矿山安全工作的快速响应和溃坝预警提供信息支撑和参考依据。
In order to solve the problem of safety early warning in the operation process of tailing pond,the case of dam failure accident of Samarco iron mine in Brazil in 2015 was selected to study the applicability of BP neural network and support vector regression( SVR) method in the prediction of water flow data. The characteristics of complexity and non-linearity of the water flow data were analyzed comprehensively,and the water flow data of tailing dam were predicted by using the above two models with taking the pond water level,rainfall and dry beach length as the input characteristics. The results showed that the maximum relative error of the prediction results based on BP neural network was not higher than 4. 35%,and the maximum relative error based on SVR algorithm was not higher than 9. 21%. The prediction results of water flow for the Fund■o dam were feasible,and the prediction accuracy of BP neural network was higher,while the calculation speed of SVR model was faster. The results can provide the information support and reference basis for the rapid response and the early warning of dam failure in the safety work of mine.
In order to solve the problem of safety early warning in the operation process of tailing pond,the case of dam failure accident of Samarco iron mine in Brazil in 2015 was selected to study the applicability of BP neural network and support vector regression( SVR) method in the prediction of water flow data. The characteristics of complexity and non-linearity of the water flow data were analyzed comprehensively,and the water flow data of tailing dam were predicted by using the above two models with taking the pond water level,rainfall and dry beach length as the input characteristics. The results showed that the maximum relative error of the prediction results based on BP neural network was not higher than 4. 35%,and the maximum relative error based on SVR algorithm was not higher than 9. 21%. The prediction results of water flow for the Fund■o dam were feasible,and the prediction accuracy of BP neural network was higher,while the calculation speed of SVR model was faster. The results can provide the information support and reference basis for the rapid response and the early warning of dam failure in the safety work of mine.
引文
[1]RICO M,BENITO G,SALGUEIRO A R,et al.Reported tailings dam failures:a review of the European incidents in the worldwide context[J].Journal of Hazardous Materials,2008,152(2):846-852.
[2]蔡嗣经,杨鹏.金属矿山尾矿问题及其综合利用与治理[J].中国工程科学,2000,2(4):89-92.CAI Sijing,YANG Peng.Tailings problems and tailings utilization and treatments in the metal mines[J].Engineering Sciences,2000,2(4):89-92.
[3]HUDSONEDWARDS K.Tackling mine wastes[J].Science,2016,352(6283):288-290.
[4]CARMO F F D,KAMINO L H Y,JUNIOR R T,et al.Fund2o tailings dam failures:the environment tragedy of the largest technological disaster of Brazilian mining in global context[J].Perspectives in Ecology&Conservation,2017,15(3):145-151.
[5]Fund2o tailings dam review panel.Report on the immediate causes of the failure of the Fund2o dam[EB/OL].http://fundaoinvestigation.com/the-panel-report,2016-8-25/2018-11-22.
[6]DONG L,SHU W,SUN D,et al.Pre-alarm system based on realtime monitoring and numerical simulation using internet of things and cloud computing for tailings dam in mines[J].IEEE Access,2017,5:21080-21089.
[7]王英博,王琳,李仲学.基于HS-BP算法的尾矿库安全评价[J].系统工程理论与实践,2012,32(11):2585-2590.WANG Yingbo,WANG Lin,LI Zhongxue.Safety evaluation of mine tailings facilities based on HS-BP algorithm[J].Systems Engineering-Theory&Practice,2012,32(11):2585-2590.
[8]李娟,李翠平,李春民,等.支持向量回归机在尾矿坝浸润线预测中的应用[J].中国安全生产科学技术,2009,5(1):76-79.LI Juan,LI Cuiping,LI Chunmin,et al.Forecasting of infiltration route in tailings dam by support vector regression[J].Journal of Safety Science and Technology,2009,5(1):76-79.
[9]TONGLE X,YINGBO W,KANG C.Tailings saturation line prediction based on genetic algorithm and BP neural network[J].Journal of Intelligent&Fuzzy Systems,2016,30(4):1947-1955.
[10]何学秋,王云海,梅国栋.基于流变-突变理论的尾矿坝溃坝机理及预警准则研究[J].中国安全科学学报,2012,22(9):74-78.HE Xueqiu,WANG Yunhai,MEI Guodong.Study on mechanism and warning criteria of tailings dam-break based on theory of rheology-mutation[J].China Safety Science Journal,2012,22(9):74-78.
[11]王昆,杨鹏,HUDSON-EDWARDS K,等.尾矿库溃坝灾害防控现状及发展[J].工程科学学报2018,40(5):526-539.WANG Kun,YANG Peng,HUDSON-EDWARDS K,et al.Current status and developing recommendations of tailings dam failure[J].Chinese Journal of Engineering,2018,40(5):526-539.
[12]张兴凯,孙恩吉,李仲学.尾矿库洪水漫顶溃坝演化规律试验研究[J].中国安全科学学报,2011,21(7):118-124.ZHANG Xingkai,SUN Enji,LI Zhongxue.Experimental study on evolution law of tailings dam flood overtopping[J].China Safety Science Journal,2011,21(7):118-124.
[13]陈凯,陆得盛,金枫,等.极端气象条件下金属矿山尾矿库在线监测系统研究[J].矿冶,2014,23(5):81-85.CHEN Kai,LU Desheng,JIN Feng,et al.Study on metallic-mine tailing dam online monitoring system in extreme weather condition[J].Min Metall,2014,23(5):81-85.
[14]CUN Y L,BOSER B,DENKER J S,et al.Handwritten digit recognition with a back-propagation network[J].Advances in Neural Information Processing Systems,1990,2(2):396-404.
[15]丁红,董文永,吴德敏.基于LM算法的双隐含层BP神经网络的水位预测[J].统计与决策,2014(15):16-19.DING Hong,DONG Wenyong,WU Demin.Water level prediction based on double hidden layer BP neural network with LM algorithm[J].Statistics&Decision,2014(15):16-19.
[16]LAW R.Back-propagation learning in improving the accuracy of neural network-based tourism demand forecasting[J].Tourism Management,2000,21(4):331-340.
[17]KAZEM A,SHARIFI E,HUSSAIN F K,et al.Support vector regression with chaos-based firefly algorithm for stock market price forecasting[J].Applied Soft Computing,2013,13(2):947-958.
[18]SCHLKOPF,BERNHARD.Learning with kernels:support vector machines,regularization,optimization,and beyond[M].Massachusetts:MIT Press,2003.
[19]赵丽娟,王慧琴,王可,等.基于多核支持向量回归的光谱反射率重建方法[J].液晶与显示,2018,33(12):1008-1018.ZHAO Lijuan,WANG Huiqin,WANG Ke,et al.Spectral reflectance reconstruction method based on multi-core support vector regression.[J].Chinese Journal of Liquid Crystals and Displays,2018,33(12):1008-1018.
[2]蔡嗣经,杨鹏.金属矿山尾矿问题及其综合利用与治理[J].中国工程科学,2000,2(4):89-92.CAI Sijing,YANG Peng.Tailings problems and tailings utilization and treatments in the metal mines[J].Engineering Sciences,2000,2(4):89-92.
[3]HUDSONEDWARDS K.Tackling mine wastes[J].Science,2016,352(6283):288-290.
[4]CARMO F F D,KAMINO L H Y,JUNIOR R T,et al.Fund2o tailings dam failures:the environment tragedy of the largest technological disaster of Brazilian mining in global context[J].Perspectives in Ecology&Conservation,2017,15(3):145-151.
[5]Fund2o tailings dam review panel.Report on the immediate causes of the failure of the Fund2o dam[EB/OL].http://fundaoinvestigation.com/the-panel-report,2016-8-25/2018-11-22.
[6]DONG L,SHU W,SUN D,et al.Pre-alarm system based on realtime monitoring and numerical simulation using internet of things and cloud computing for tailings dam in mines[J].IEEE Access,2017,5:21080-21089.
[7]王英博,王琳,李仲学.基于HS-BP算法的尾矿库安全评价[J].系统工程理论与实践,2012,32(11):2585-2590.WANG Yingbo,WANG Lin,LI Zhongxue.Safety evaluation of mine tailings facilities based on HS-BP algorithm[J].Systems Engineering-Theory&Practice,2012,32(11):2585-2590.
[8]李娟,李翠平,李春民,等.支持向量回归机在尾矿坝浸润线预测中的应用[J].中国安全生产科学技术,2009,5(1):76-79.LI Juan,LI Cuiping,LI Chunmin,et al.Forecasting of infiltration route in tailings dam by support vector regression[J].Journal of Safety Science and Technology,2009,5(1):76-79.
[9]TONGLE X,YINGBO W,KANG C.Tailings saturation line prediction based on genetic algorithm and BP neural network[J].Journal of Intelligent&Fuzzy Systems,2016,30(4):1947-1955.
[10]何学秋,王云海,梅国栋.基于流变-突变理论的尾矿坝溃坝机理及预警准则研究[J].中国安全科学学报,2012,22(9):74-78.HE Xueqiu,WANG Yunhai,MEI Guodong.Study on mechanism and warning criteria of tailings dam-break based on theory of rheology-mutation[J].China Safety Science Journal,2012,22(9):74-78.
[11]王昆,杨鹏,HUDSON-EDWARDS K,等.尾矿库溃坝灾害防控现状及发展[J].工程科学学报2018,40(5):526-539.WANG Kun,YANG Peng,HUDSON-EDWARDS K,et al.Current status and developing recommendations of tailings dam failure[J].Chinese Journal of Engineering,2018,40(5):526-539.
[12]张兴凯,孙恩吉,李仲学.尾矿库洪水漫顶溃坝演化规律试验研究[J].中国安全科学学报,2011,21(7):118-124.ZHANG Xingkai,SUN Enji,LI Zhongxue.Experimental study on evolution law of tailings dam flood overtopping[J].China Safety Science Journal,2011,21(7):118-124.
[13]陈凯,陆得盛,金枫,等.极端气象条件下金属矿山尾矿库在线监测系统研究[J].矿冶,2014,23(5):81-85.CHEN Kai,LU Desheng,JIN Feng,et al.Study on metallic-mine tailing dam online monitoring system in extreme weather condition[J].Min Metall,2014,23(5):81-85.
[14]CUN Y L,BOSER B,DENKER J S,et al.Handwritten digit recognition with a back-propagation network[J].Advances in Neural Information Processing Systems,1990,2(2):396-404.
[15]丁红,董文永,吴德敏.基于LM算法的双隐含层BP神经网络的水位预测[J].统计与决策,2014(15):16-19.DING Hong,DONG Wenyong,WU Demin.Water level prediction based on double hidden layer BP neural network with LM algorithm[J].Statistics&Decision,2014(15):16-19.
[16]LAW R.Back-propagation learning in improving the accuracy of neural network-based tourism demand forecasting[J].Tourism Management,2000,21(4):331-340.
[17]KAZEM A,SHARIFI E,HUSSAIN F K,et al.Support vector regression with chaos-based firefly algorithm for stock market price forecasting[J].Applied Soft Computing,2013,13(2):947-958.
[18]SCHLKOPF,BERNHARD.Learning with kernels:support vector machines,regularization,optimization,and beyond[M].Massachusetts:MIT Press,2003.
[19]赵丽娟,王慧琴,王可,等.基于多核支持向量回归的光谱反射率重建方法[J].液晶与显示,2018,33(12):1008-1018.ZHAO Lijuan,WANG Huiqin,WANG Ke,et al.Spectral reflectance reconstruction method based on multi-core support vector regression.[J].Chinese Journal of Liquid Crystals and Displays,2018,33(12):1008-1018.