基于WSN的支架初撑力提取与工作阻力预测方法研究

发布时间：2018-03-06 11:03

本文选题：支架　切入点：无线传感器网络　出处：《中国矿业大学》2015年硕士论文　论文类型：学位论文

【摘要】：不断增大的煤矿采高对液压支架的支护能力提出了更高的要求。初撑力和来压值是液压支架最重要的两个指标,对煤矿安全开采具有现实指导意义。目前,液压支架支设的初撑力和来压值基本由支架操作人员主观经验确定,使得很多支架在实际工作时的初撑力低于其设计值。因此本课题设计了一套基于无线传感器网络(Wireless Sensor Networks,WSN)的工作阻力监测系统;通过对支架工作阻力曲线的分析,提出了一种支架初撑力和周期来压值的自动提取方法;并利用极限学习理论实现了支架工作阻力的短期预测。研究内容主要包括:首先,给出了基于WSN的支架工作阻力监测系统的软件设计方法。以Microsoft Visual Studio 2008、Microsoft SQL Server 2005为集成开发环境,利用C#和SQL语言实现了数据的接收、存储、显示、查询、曲线绘制等功能。其次,提出了支架初撑力提取方法。首先,借鉴非均匀量化思想,利用A率压缩算法对液压支架工作阻力曲线进行非均匀量化。因为初撑力区的数据变化剧烈,必然会产生最多的量化段数。采用滑动窗口法找到量化段数最多的初撑力区。然后利用能量比法在初撑力区提取出初撑力值。最后将本文方法提取的初撑力值和经验估算值进行了对比,验证了算法的有效性。最后,利用极限学习机(Extreme Learning Machine,ELM)理论把工作阻力历史数据作为ELM的训练集样本。通过分析隐含层神经元个数对ELM性能的影响,本次实验中将隐含层节点数定为20来确定ELM网络的模式。通过实验仿真,预测曲线与实际曲线拟合效果良好,除了在移架过程中的预测误差超过了10%,预测输出误差率都在2%附近。
[Abstract]:The increasing mining height of coal mine has put forward higher requirements for the support ability of the hydraulic support. The initial support force and the pressure value are the two most important indexes of the hydraulic support, which have practical guiding significance for the safe mining of coal mine. The initial support force and pressure value of the hydraulic support are basically determined by the subjective experience of the support operator. This paper designs a set of working resistance monitoring system based on Wireless Sensor Networks (WSNs) of wireless sensor network, and analyzes the working resistance curve of the support. In this paper, an automatic method for extracting the initial support force and periodic pressure of the support is proposed, and the short-term prediction of the support working resistance is realized by using the limit learning theory. The main contents of the research are as follows: first of all, The software design method of support working resistance monitoring system based on WSN is presented. Taking Microsoft Visual Studio 2008 Microsoft SQL Server 2005 as the integrated development environment, the functions of data receiving, storing, displaying, querying, curve drawing and so on are realized by using C # and SQL language. In this paper, a method of extracting support initial support force is proposed. Firstly, using the idea of non-uniform quantization and A rate compression algorithm for non-uniform quantization of the working resistance curve of hydraulic support, because the data of the initial bracing force region change dramatically, The sliding window method is used to find the first bracing force region with the largest number of quantized segments. Then the initial bracing force is extracted from the initial bracing force region by using the energy ratio method. Finally, the sum of the initial bracing force values obtained by the method in this paper is given. The empirical estimates are compared, The validity of the algorithm is verified. Finally, the working resistance history data are taken as the training set samples of ELM by using extreme Learning Machine (ELM) theory. The effect of the number of hidden neurons on the performance of ELM is analyzed. In this experiment, the number of hidden layers is set at 20 to determine the model of ELM network. The simulation results show that the predicted curve fits well with the actual curve. With the exception of more than 10 prediction errors in the moving frame, the predicted output error rates are around 2%.
【学位授予单位】：中国矿业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TD355.4

【参考文献】