基于支持向量机的尾矿库风险预测算法研究
发布时间：2019-02-25 12:00
【摘要】：随着工业化经济的迅速发展,社会对矿山资源的需求总量日渐增加;但由于矿石开采率不高以及尾矿库建造低成本导致我国90%以上尾矿处置采用尾矿库堆积的方式,逐渐形成了尾矿库数量庞大、安监人员不足的局面。近年,尾矿库坝体垮坍导致的群众伤亡事故频发,对社会群众的财产和人身安全造成了严重危害。所以提出一种科学精准、快捷可靠的尾矿库安全风险评估预测方法来满足政府及企业对尾矿库的监管需求具有重要意义。现有的尾矿库风险评估模型大多缺乏对尾矿库空间指标评估,并且对所建立的尾矿库监测系统采集的数据中隐藏信息和内在关联尚未进行充分挖掘。针对上述问题,本文提出一种基于改进支持向量机的尾矿库安全风险预测方法。全文的主要工作和主要成果如下:(1)分析尾矿库风险评估方法研究现状,总结了尾矿库风险评估领域不同理论方法的局限性;本文增加了尾矿指标空间位置信息,并建立尾矿库风险时空评价模型,完善了尾矿库风险评估在空间信息上的不足。(2)依据尾矿库工程结构,分析尾矿风险事故的主要原因,根据干滩长度、库水位、坝体位移、地下位移等监测指标的特点,通过特征提取明确风险评估中的关键监测指标。(3)将集成学习思想应用于支持向量机回归,构建一种基于堆栈(Stacking)学习集成支持向量机预测算法进行尾矿库风险等级预测,开展详细仿真实验,结果表明改进后的集成支持向量机在训练速度、预测准确率等方面具有一定的优越性。(4)将一种改进的新型元启发式布谷鸟搜索算法,对支持向量机进行寻优得到全局最优解,从而得到具有最佳参数的支持向量机尾矿库风险评估预测模型。并将改进的尾矿库风险预测算法与粒子群算法、遗传算法和原始布谷鸟算法在运用中的结果对比,证明其算法改进得到的模型拟合准确率更好。本文成果验证了改进后的支持向量机算法是可行的尾矿库风险等级预测方法,能准确预测尾矿库风险等级,对保障尾矿库安全与预警工作提供了新的思路和理论方法。
[Abstract]:With the rapid development of industrialized economy, the total demand for mining resources is increasing day by day. However, due to the low mining rate and the low cost of tailing reservoir construction, more than 90% of tailing disposal in our country adopts the way of tailings deposit accumulation, which gradually forms the situation of huge quantity of tailing reservoir and shortage of safety supervision personnel. In recent years, accidents of mass casualties caused by collapse of tailing dam have caused serious harm to the property and personal safety of the masses. Therefore, it is of great significance to put forward a scientific, accurate, fast and reliable method to evaluate and forecast the safety risk of tailings reservoir to meet the needs of the government and enterprises. Most of the existing tailing reservoir risk assessment models lack the evaluation of tailing reservoir spatial indicators, and the hidden information and internal correlation in the data collected by the tailing reservoir monitoring system have not been fully excavated. In order to solve the above problems, this paper presents a method for predicting the safety risk of tailings reservoir based on improved support vector machine (SVM). The main work and results are as follows: (1) analyzing the present situation of tailing reservoir risk assessment methods and summarizing the limitations of different theoretical methods in tailing pool risk assessment field; In this paper, the spatial location information of tailing index is added, and the spatio-temporal evaluation model of tailing reservoir risk is established, which improves the deficiency of spatial information of tailing reservoir risk assessment. (2) according to the structure of tailing reservoir engineering, The main causes of risk accidents of tailings are analyzed. According to the characteristics of monitoring indexes such as dry beach length, reservoir water level, dam body displacement, underground displacement, etc. The key monitoring indicators in risk assessment are identified by feature extraction. (3) the integrated learning idea is applied to support vector machine regression. A prediction algorithm based on stack (Stacking) learning integrated support vector machine is constructed to predict the risk level of tailings reservoir. The simulation results show that the improved integrated support vector machine is training speed. Prediction accuracy has some advantages. (4) an improved meta-heuristic Cuckoo search algorithm is used to optimize the support vector machine (SVM) to obtain the global optimal solution. The risk assessment and prediction model of tailings reservoir based on support vector machine (SVM) with optimal parameters is obtained. The improved tailing reservoir risk prediction algorithm is compared with particle swarm algorithm, genetic algorithm and original cuckoo algorithm in the application, and the improved model fitting accuracy is proved to be better than that of the original Cuckoo algorithm. The results of this paper verify that the improved support vector machine algorithm is a feasible method for predicting the risk grade of tailing reservoir, which can accurately predict the risk grade of tailing reservoir, and provides a new thinking and theoretical method for ensuring the safety and early warning of tailing reservoir.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD926.4;TP18
[Abstract]:With the rapid development of industrialized economy, the total demand for mining resources is increasing day by day. However, due to the low mining rate and the low cost of tailing reservoir construction, more than 90% of tailing disposal in our country adopts the way of tailings deposit accumulation, which gradually forms the situation of huge quantity of tailing reservoir and shortage of safety supervision personnel. In recent years, accidents of mass casualties caused by collapse of tailing dam have caused serious harm to the property and personal safety of the masses. Therefore, it is of great significance to put forward a scientific, accurate, fast and reliable method to evaluate and forecast the safety risk of tailings reservoir to meet the needs of the government and enterprises. Most of the existing tailing reservoir risk assessment models lack the evaluation of tailing reservoir spatial indicators, and the hidden information and internal correlation in the data collected by the tailing reservoir monitoring system have not been fully excavated. In order to solve the above problems, this paper presents a method for predicting the safety risk of tailings reservoir based on improved support vector machine (SVM). The main work and results are as follows: (1) analyzing the present situation of tailing reservoir risk assessment methods and summarizing the limitations of different theoretical methods in tailing pool risk assessment field; In this paper, the spatial location information of tailing index is added, and the spatio-temporal evaluation model of tailing reservoir risk is established, which improves the deficiency of spatial information of tailing reservoir risk assessment. (2) according to the structure of tailing reservoir engineering, The main causes of risk accidents of tailings are analyzed. According to the characteristics of monitoring indexes such as dry beach length, reservoir water level, dam body displacement, underground displacement, etc. The key monitoring indicators in risk assessment are identified by feature extraction. (3) the integrated learning idea is applied to support vector machine regression. A prediction algorithm based on stack (Stacking) learning integrated support vector machine is constructed to predict the risk level of tailings reservoir. The simulation results show that the improved integrated support vector machine is training speed. Prediction accuracy has some advantages. (4) an improved meta-heuristic Cuckoo search algorithm is used to optimize the support vector machine (SVM) to obtain the global optimal solution. The risk assessment and prediction model of tailings reservoir based on support vector machine (SVM) with optimal parameters is obtained. The improved tailing reservoir risk prediction algorithm is compared with particle swarm algorithm, genetic algorithm and original cuckoo algorithm in the application, and the improved model fitting accuracy is proved to be better than that of the original Cuckoo algorithm. The results of this paper verify that the improved support vector machine algorithm is a feasible method for predicting the risk grade of tailing reservoir, which can accurately predict the risk grade of tailing reservoir, and provides a new thinking and theoretical method for ensuring the safety and early warning of tailing reservoir.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD926.4;TP18
【参考文献】
相关期刊论文 前10条
1 王英博;聂娜娜;王铭泽;李仲学;;修正型果蝇算法优化GRNN网络的尾矿库安全预测[J];计算机工程;2015年04期
2 陆志强;李吉鹏;章耕耘;马丽;;基于可变模糊评价模型的东山湾生态系统健康评价[J];生态学报;2015年14期
3 刘传立;刘小生;李妍妍;;基于GEP的金属矿尾矿坝变形预测模型研究[J];有色金属科学与工程;2013年06期
4 杨小彬;李和明;尹忠东;蒋利民;孟s，
本文编号：2430141
本文链接：https://www.wllwen.com/kejilunwen/kuangye/2430141.html
