基于PCA-MSA的矿井突水水源判别算法研究

发布时间：2018-04-10 21:24

本文选题：矿井突水 + 主成分分析法　；参考：《中南大学》2013年硕士论文

【摘要】：矿井突水是威胁矿山安全生产的最大自然灾害之一,正确识别突水水源可为制定合适的防治水措施提供重要依据,因此,如何快速有效的判别矿井突水水源具有重要意义。目前,大部分水源判别模型没有考虑到由于水化学判别指标之间信息叠加从而导致水源误判的影响。为此,本文以某两个矿山(A、B矿山)的水文地质资料和水化学数据研究对象,引用了主成分分析法对水化学指标进行信息提炼,并结合两种多元判别方法来建立水源判别模型,并与传统的多元判别模型(Fisher判别法、Bayes判别法、逐步判别法)进行比较,找出最优的水源判别模型。同时,为了让矿井突水水源判别工作更加自动化和智能化,通过C++语言开发了矿井突水水源判别系统。本文研究主要结果如下： (1)选取A矿山各水层的Ca2+、Mg2+、K++Na+、HCO3-、SO42-Cl-、TDS的7种水化学指标参数为数据样本,利用Aquachem等水化学分析软件绘制出Piper三线图、常规离子箱图、常规离子与TDS的关系图等来分析各水层水化学特性。结果表明各水源具有不同的水化学特性,且太灰水与奥灰水的各离子分布范围比较接近,两者水化学特性较相似,难以通过某一单个离子分布规律来区分突水水源。 (2)以A、B矿山的水化学信息为研究对象,建立Fisher判别模型、Bayes判别模型、逐步判别模型、和与之相应的经过主成分分析的两种判别模型。研究结果表明,经过主成分分析处理后的Fisher判别模型、Bayes判别模型A矿山的回判和预判正确率分别得到提高了6%、18%,完全优于传统的Fisher判别模型、Bayes判别模型。通过主成分分析处理后消除了因为各指标信息的叠加而导致其水源误判的问题,提高了模型识别突水的精度。 (3)以矿区各水源的水化学的差异性为理论基础,以Fisher判别法、Bayes判别法和主成分分析法相结合的模型为核心,采用C++语言开发出矿井突水水源判别系统。该系统不仅具有数据存储、查询功能,还可以快速的判别突水水源,并具有良好的实用性和可靠性,操作简便。
[Abstract]:Mine water inrush is one of the biggest natural disasters threatening mine safety production. The correct identification of water inrush source can provide an important basis for making appropriate water prevention measures. Therefore, how to quickly and effectively distinguish mine water inrush source is of great significance.At present, most water source discriminant models do not take into account the influence of water source misjudgment due to the superposition of information among the water chemistry discriminant indexes.Therefore, based on the hydrogeological data and hydrochemical data of two mines, the principal component analysis (PCA) is used to extract the information of the hydrochemical indexes, and two multivariate discriminant methods are used to establish the water source discriminant model.Compared with the traditional multivariate discriminant model, Fisher discriminant method, Bayes discriminant method and stepwise discriminant method, the optimal water source discriminant model is found out.At the same time, in order to make the discrimination of mine water inrush water source more automatic and intelligent, a mine water inrush water source discrimination system is developed by C language.The main results of this paper are as follows:1) 7 kinds of hydrochemical index parameters of Ca2 mg _ 2 K _ 2O H _ CO _ 3-so _ 42-Cl-TDs in each water layer of mine A are selected as data samples, and Piper three-line diagram and conventional ion box diagram are drawn by using Aquachem isohydrochemical analysis software.The relationship between conventional ions and TDS was used to analyze the hydrochemical characteristics of each water layer.The results show that the water sources have different hydrochemical characteristics, and the distribution range of the ions in the ash water is similar to that in the Ordovician water, so it is difficult to distinguish the water inrush water sources by a single ion distribution law.Through principal component analysis, the problem of water source misjudgment caused by the superposition of index information is eliminated, and the accuracy of water inrush recognition is improved.3) based on the differences of water chemistry of water sources in mining areas, and based on the model of combining Fisher discriminant method with principal component analysis method, a discriminant system of mine water inrush water source is developed by C language.The system not only has the functions of data storage and query, but also can quickly distinguish the water source of water inrush, and has good practicability, reliability and simple operation.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TD745

【参考文献】