Hadoop平台上煤矿企业储备定额算法并行化研究与应用

发布时间：2018-02-21 12:06

本文关键词： 备件消耗量预测概率统计分析法 MapReduce 模糊综合评价法　出处：《内蒙古科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：某煤矿集团公司成功的引进了SAP公司的ERP（Enterprise Resource Planning）系统，ERP系统的使用给企业带来了先进的管理理念，建设完成了完整的企业资原管理体系和高效、便捷的信息技术平台。但是，，上述ERP系统分析和计算备件的储备定额侧重于机械制造等备件消耗规律性较强的行业，对于煤矿企业这类备件消耗随需求变化的行业起不到应有的作用，所以开发了备件储备定额系统来对备件信息进行管理，协助业务人员制定备件采购计划，自动提示所需订货的备件等。但是，随着系统的使用，一些问题也紧跟着暴露出来。如储备定额系统对于日常少量备件做消耗量预测可以在较短的时间内很好的完成，可在年中需要为下半年做订购计划或为来年制定订购计划的时候，因为其备件库的庞大（现常用备件有29万多种，历史出入库存记录数据更多），做消耗量预测需要花费很长的时间；另外备件储备定额模型中缓冲存储量计算中用到了裕度系数，储备定额系统中裕度系数的选取是采购人员人为确定的，没有采用科学的方法从订货周期内的生产计划、往年同期的备件消耗量、备件的供应情况等因素综合确定，这样的后果是主观因素大，影响准确性。因此怎样有效的计算出预测值制订订货计划，以及提出一个裕度系数的确定方法是本文所要解决的问题。随着Hadoop云计算平台在各个领域的运用很好的证明了其对海量数据的存储能力和并行计算能力，这为解决大量备件的消耗量预测提供了一种新的解决方式，本文提出基于Hadoop云计算平台的备件消耗量预测系统。该系统分为数据获取模块、数据存储模块、数据预处理模块和备件消耗量预测模块四部分。其中，数据获取模块利用某煤炭集团公司的ERP系统Web Service接口来获取用户数据；数据存储模块中将数据获取模块中获取的备件数据按设计的数据格式存入本地Oracle数据库中；数据预处理模块利用VS2010开发程序将数据库中备件数据按要求进行处理，得到我们需要的数据格式的数据，通过多层次模糊综合评价法，从备件的关键性和备件所属设备的关键性两方面对备件重要性进行评价，量化备件重要性得到所需的裕度系数K；备件消耗量预测模块中对备件消耗量预测方法（概率统计分析法）进行改进，以经典矩阵相乘的经典算法为基础，利用MapReduce编程框架进行MapReduce化设计，构建MapReduce并行处理算法并在MapReduce并行编程模型上实现。实验结果表明，经过MapReduce设计的算法在处理器的可扩展性、数据的可扩展性和加速比性能这三方面的实验中具有良好的指标，算法性能表现良好。
[Abstract]:A coal mine group company successfully introduced the ERP(Enterprise Resource planning system of SAP Company to bring the advanced management idea to the enterprise, and completed the complete enterprise capital original management system and the efficient and convenient information technology platform. The above ERP system analysis and calculation of spare parts reserve quota is focused on the industries with strong regularity of spare parts consumption, such as mechanical manufacturing, which does not play a due role in the industries where the consumption of spare parts varies with the demand of coal mining enterprises. So the spare parts reserve quota system has been developed to manage the spare parts information, to assist the business personnel to draw up the spare parts purchase plan, to automatically prompt the spare parts that need to be ordered, etc. However, with the use of the system, Some problems have come to light. For example, the reserve quota system can predict the consumption of a small amount of spare parts in a short time. When you need an order plan for the second half of the year in the middle of the year, or an order plan for the coming year, because of the size of its spare parts warehouse (there are now more than 290,000 commonly used spare parts, In addition, the margin coefficient is used in the calculation of buffer storage capacity in spare parts reserve quota model. The selection of margin coefficient in the reserve quota system is artificially determined by the purchasing personnel, and no scientific method is used to determine the production plan in the order cycle, the consumption of spare parts in the same period of previous years, the supply of spare parts, and so on. Therefore, how to calculate the forecast value effectively to make the order plan, and to put forward a method of determining the margin coefficient is the problem to be solved in this paper. With the application of Hadoop cloud computing platform in various fields, it has proved its storage capacity and parallel computing ability of massive data, which provides a new solution to solve the consumption prediction of a large number of spare parts. This paper presents a spare parts consumption prediction system based on Hadoop cloud computing platform. The system is divided into four parts: data acquisition module, data storage module, data preprocessing module and spare parts consumption prediction module. The data acquisition module uses the Web Service interface of the ERP system of a coal group company to obtain the user data, and the data storage module stores the spare parts data obtained from the data acquisition module into the local Oracle database according to the designed data format. The data preprocessing module uses the VS2010 development program to process the spare parts data in the database according to the requirement, and obtains the data of the data format that we need, through the multi-level fuzzy comprehensive evaluation method, The importance of spare parts is evaluated in terms of the criticality of the spare parts and the criticality of the equipment to which the spare parts belong, The margin coefficient K is obtained by quantifying the importance of spare parts, and the prediction method of spare parts consumption (probabilistic and statistical analysis) is improved in the prediction module of spare parts consumption, which is based on the classical algorithm of multiplying the classical matrix. The MapReduce parallel processing algorithm is constructed and implemented on the MapReduce parallel programming model using the MapReduce programming framework. The experimental results show that the algorithm designed by MapReduce is extensible in the processor. The experiments of data scalability and speedup performance have a good performance, and the performance of the algorithm is good.
【学位授予单位】：内蒙古科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP311.13;F224;F426.21

【参考文献】