深度神经网络算法在尾矿库安全评价中的应用研究

发布时间：2018-04-14 14:29

本文选题：尾矿库 + 安全评价　；参考：《浙江工业大学》2015年硕士论文

【摘要】：当下的经济和工业快速发展使得矿产资源的需求快速增加,同时尾矿库的数量也随之增长。由于运行期间坝高和库容都会上升,尾矿库逐渐变为潜在的高势能危险源,对矿山和选矿厂本身的安全运营、下游群众的生命财产安全以及周边自然环境都构成了巨大威胁。因此设计一个高效、合理的尾矿库安全评价方法,对尾矿安全的趋势作出准确预测具有重大现实意义。为了充分挖掘影响尾矿库安全的因素中的隐藏信息和内在联系,获得尽可能高的安全预测准确率,本文设计了一种基于深度神经网络的尾矿库安全趋势预测方法。全文的主要工作和主要成果如下:1)总结尾矿库安全评价方法的研究现状,分析本领域目前为止所使用的理论和方法,对它们的优点和缺点进行了比较。2)介绍了深度神经网络原理,阐述了堆栈式自编码器的设计思路和算法,包括深度神经网络的逐层优化训练思想和用于参数训练LM-BP算法等。3)根据尾矿库的事故原因统计分析和其工程结构特点,分析导致尾矿库事故的主要原因,给出了尾矿库安全关键因素。4)应用深度神经网络进行尾矿库安全趋势预测实验,给出了详细仿真结果分析。结果表明深度神经网络在特征抽象、表征学习、预测准确率等方面都具有优越性。5)给出了尾矿库安全评价软件的设计思路,并使用PYTHON及相关的软件包实现了软件原型,以便算法得到实际应用。本文成果验证了深度神经网络是可行的安全状态趋势预测方法,为尾矿库安全评价工作提供了新的思路和理论支持。
[Abstract]:With the rapid development of economy and industry, the demand for mineral resources is increasing rapidly, and the number of tailings is also increasing.Due to the rise of dam height and reservoir capacity during operation, the tailing reservoir gradually becomes a potential high potential energy hazard source, which poses a great threat to the safe operation of mine and concentrator itself, the safety of life and property of downstream people and the surrounding natural environment.Therefore, it is of great practical significance to design an efficient and reasonable method to evaluate the safety of tailings.In order to fully excavate the hidden information and internal relation of the factors affecting the safety of tailings reservoir and obtain the highest accuracy of safety prediction, a method of predicting the safety trend of tailing reservoir based on depth neural network is designed in this paper.The main work and main results of this paper are as follows: 1) summarizing the research status of tailing reservoir safety evaluation methods, analyzing the theories and methods used so far in this field.The principle of depth neural network is introduced, and the design idea and algorithm of stack self-encoder are described.Including the depth neural network layer by layer optimization training idea and LM-BP algorithm for parameter training, etc. 3) according to the statistical analysis of the accident cause of tailing reservoir and its engineering structure characteristics, the main reasons leading to the tailing reservoir accident are analyzed.The key factor of tailing reservoir safety. 4) the experiment of predicting the safety trend of tailing reservoir by using depth neural network is presented, and the detailed simulation results are given.The results show that the depth neural network has advantages in feature abstraction, representation learning and prediction accuracy. (5) the design idea of safety evaluation software for tailings reservoir is given, and the software prototype is realized by using PYTHON and related software packages.So that the algorithm can be applied in practice.The results of this paper verify that the depth neural network is a feasible method for predicting the trend of safety state, and provides a new way of thinking and theoretical support for the safety evaluation of tailings 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】：2015
【分类号】：TD926.4;TP183

【共引文献】