节点文献

神经网络系统开发及巷道变形研究

Neural Network System Development and Research of Laneway Distortion

【作者】 尹潘

【导师】 盛建龙;

【作者基本信息】 武汉科技大学 , 采矿工程, 2003, 硕士

【摘要】 余华寺矿是武钢矿业公司的一个矿山,近年来,在-137.5m和-150m两个分段中的一些开采巷道出现较大的变形,并直接影响到该矿山的生产和施工。本论文对余华寺矿巷道的变形因素进行分析,通过对不同时间的实测巷道变形值进行神经网络学习,把学习出的结果作为检验依据,来对巷道一段时间后的变形进行预测,并取得了较好的结果。 作者开发出一套BP神经网络计算软件——BPNNS系统,它是基于Windows 2000 Server平台的C/S构架,能实现远程访问和控制。编码是由Inprise公司开发出的Delphi 6.0来实现的。系统友好的界面便于计算过程的控制和即时信息修改、管理和查询;信息的存储是由微软公司开发出的SQL Server 2000来实现的,它能存储大量数据,并用自带的数据库语言Transact—SQL来实现对存储信息的存取、管理、修改和查询。 在论文期间,作者进行了大量的现场观测、实验和数据采集,参考了大量文献和资料,并开发出了适用性较广的BPNNS系统,取得了以下成果和认识: 1)参与了武钢余华寺矿采场地压规律分析及控制对策研究项目。对该矿山的地压分布规律进行研究,得出其规律,并用数值模拟来对该地压下的回采顺序进行研究。 2)对变形巷道的地压进行研究。对该变形巷道的地压分布规律进行研究,分析地压显现的基本原理,提出影响地压分布的因素,包括一些自然因素和人为因素,初步提出了地压分布对采场稳定性的影响。 3)分析影响巷道变形可能的因素。作者对这些因素进行分析,把有重要影响的因素用于预测巷道变形的计算。 4)用模糊算法——BP神经网络算法来实现对巷道变形的预测。由于影响巷道变形的因素很多,有些因素我们了解很多,有些了解很少,对于一些突发因素的影响,则更有不可预见性。并且这些因素之间相互影响,因而需要用模糊计算来解决这种问题,神经网络算法用其极强的非线性动态处理能力来对巷道的变形进行预测。从计算的结果来看,用神经网络算法来处理这类问题是比较合适的。 5)探讨并实现用Delphi 6.0对神经网络算法进行编码。由于在神经网络计算过程中需要信息和数据的输入和输出、调整参数、计算条件选择、计算过程控制和信息查询,这需要有一个比较友好的界面来实现这些功能。Delphi以其友好界面、强大的数据库功能成功地用于神经网络计算中。 6)用SQL Server 2000探讨并实现神经网络计算和工程信息、数据的存储、输出、管理、修改和查询功能。SQL Server 2000能存储大量的数据,其存储信息量几乎能达到硬盘的容量,其自身带的数据库编程语言Transact-SQL能方便地对数据库、数据表和表中的各字段进行各种操作。BPNNS系统中几乎所有神经网络计算和工程项目信息都是存储在数据库中的,它为网络计算的工程信息查询提供了数据和信息。 7)开发出一套BP神经网络算法软件——BPNNS系统。它有以下特点: (1)它能适用于所有的BP神经网络算法,并且不受网络结构的限制,对于网络结构的不同层数和各层的不同节点数,系统都能根据该工程创建时所输入的信息动态地创建计算所需要的各种控件,接受系统输入,并把输入结果存储到数据库中。 (2)它有新增、修改、查询和删除四大功能组成。对于新的工程需要创 建;对于计算结果不合理的网络结构要对它进行修改:当查看该工程的数 据和信息时,需要对该工程进行查询;如果不想在系统中保留无用的工程 项目时,就可以对工程进行删除。 (3)当所选定的网络结构不合理时,可以方便的修改网络结构,然后再 计算,而不用重新构建网络结构。修改的过程,就是对创建工程时输入的 信息进行修改,并新增相应的数据库和数据表信息,修改后的计算过程不 变。 (4)改变了神经网络算法没有“记忆性”的特点。系统能够记录下网 络中的权值和闽值的变化,为分析项目的影响因素提供参考。虽然权值和 闽值的变化过程对网络的检测没有实际作用,但我们可以看出合理的网络 结构的权值和阐值的变化过程。 (5)系统大量使用图形来形象地展现误差的变化。我们选择合理的网络 结构主要是依据误差大小和波动情况,系统能把每次的计算误差用点图和 线图的形式表现出来,还能把一个工程中不同网络结构的误差在一起比较, 从而选择出合理的网络结构。 (6)系统尽可能的提供用户接口,来实现使用者对参数选择和计算条件 的控制。 (7)系统提供大量的信息查询,工程的所有信息和数据都能以数据表的 形式查询出来。对于系统使用者信息、工程信息和系统使用信息,系统则 以报表的形式形象地展示出来。 总之,对巷道的变形因素进行分析,用神经网络算法进行计算,利用BPNNS系统能够方便、高效地进行巷道变形预测,并取得较好的结果。

【Abstract】 Yuhuasi Mine is one of the mines which belong to Mine Company of Wugan Steal and Iron Group. But recently, there is great distortion in laneways of phases of -137.5m and -150m, which have directly affected the production and structure of the mine. In this thesis, the author analyzed the factors which affected the distortion of the laneways. Through the neural network study of the data of distortion in different time, the result of the study is looked as the basis of checking to predict the distortion and gain good result.The author exploited a set of neural network Algorithm software--BPNNS System, which is C/S truss and based on the operation system of Windows 2000 Server. So it can be accessed and controlled remotely. The system is coded in Delphi 6.0 which is exploited by Inprise Company. The friendly forms are useful to control the process, change information in real time, manage, and query. SQL Server 2000, exploited by Software Company, stores the information. Because of the its database language --Transact-SQL, we can store, manage, change and query easily.During the period of the thesis, the author did lots of work to observe in the port, experiment, gather data, reference lots of documents and data, and exploited a useful BPNNS System, and gained the following achievements.1) Attending the researching item of pressure analysis and control of Yuhuasi Mine. The author analyzed the pressure distributing rules of the mine and gained the rules, researched the order of mining using digital simulation.2) Researching the pressure of distorted laneways. The author analyzed the distributing rules of pressure of laneway, analyzed the fundamental principle, brought forward the affected factors of pressure, including some natural factors and manual factors, and brought forward the effect, caused by pressure, to the stability of mine primarily.3) Listing the possible factors affecting the laneway distortion. The author used some important factors to predict the distortion.4) Using fuzzy algorithm--BP neural network to predict the lanewaydistortion. There are many factors affected the distortion. Some of them we know a lot. Some of them we know very little. And the factors affect each other. So we need fuzzy algorithm to solve the problem. The neural network algorithm can predict the distortion because of its powerful capability of non-linearity and dynamic disposal. From the calculated result, we know the algorithm is suit to solve the problem.5) Researching and realization the coding using Delphi 6.0. Because it is needed in the process to input and output information and data, change parameters, choose calculated situation, control the process of calculation and query information. All of above need friendly forms.Delphi 6. 0 is used in the algorithm successfully because of its friendly forms and powerful database capability.6) Researching and realization the neural network calculation, the store of project information and data, output, management, changing, query using SQL Sever 2000. SQL Server 2000 can store a great deal of data, even reach to the capacity of hard disk. Its coding language--Transact-SQL, can operate databases, tables and fields easily. In BPNNS System, all information of projects and calculation are stored in the database, which provides the data and information for the query.7) Exploit a set of BP neural network software--BPNNS, whosecharacters are as follows:(1) It is suit for all kinds of BP neural network algorithm, and not limited by the network structure. The system can create amount of textbox dynamically to accept input and store it to database.(2) The system has 4 main functions, such as adding, changing, query and deletion. We create new project and change the structure of network when necessary. We also can query the information of system and project and delete the useless project.(3) If the network structure is not reasonable, we have to change it and recalculate instead of creating new network structure. During the changin

  • 【分类号】TD322
  • 【被引频次】1
  • 【下载频次】155
节点文献中: 

本文链接的文献网络图示:

本文的引文网络