基于云计算的设备故障趋势预测方法研究

发布时间：2018-11-27 07:47

【摘要】：现代化工业发展过程中,庞大的数据已经成为企业关注的重要资源,如何安全保存、共享企业大数据资源,挖掘其中潜藏的价值,亟待深入研究。云计算提出了基础设施即服务、平台即服务、软件即服务的全新服务模式,适应企业不同阶段的需求,为现代化工业提出了一种全新的发展模式。本文将云计算技术与传统设备维护系统相结合,提出了基于Hadoop的设备维护系统云平台,设备维护分布式文件系统,设备维护分布式计算框架,并分别从设备维护资源层、设备维护服务层、设备维护应用层三个层面对设备维护云平台系统进行了详细的论述。重点研究了设备维护服务层的设备故障趋势预测模块,采用支持向量回归机算法进行故障趋势预测,同时分析了参数C、8、θ对支持向量回归机性能产生的不同影响,通过粒子群优化算法对支持向量回归机进行参数优化。采用UCI数据库中的一组标准数据集进行了优化实验。实际应用中,数据规模逐渐走向巨大化,传统支持向量回归机所需要的时间急剧增加,针对这一问题,提出基于Hadoop环境下分布式支持向量回归机算法。实验研究表明,基于Hadoop分布式支持向量回归机与传统支持向量回归机在预测性能基本持平的基础上,大大节省了计算时间。同时,分析了在保持样本数据不变的情况下,增加Map任务数量对时间消耗的影响,得出在一定范围内增加Map任务数量会减少时间消耗。建立了基于Hadoop分布式支持向量回归机的设备故障趋势预测模型,利用某煤炭企业采集的设备振动数据,对该模型的预测性能进行了验证。结果表明,基于Hadoop分布式支持向量回归机在故障趋势预测中具有节约时间、预测精度高、可靠性好等特点,能够满足实际使用要求。
[Abstract]:In the process of modern industrial development, huge data has become an important resource for enterprises to pay close attention to. How to save and share big data resources of enterprises safely and excavate the hidden value of them is urgent to be deeply studied. Cloud computing proposes a new service model of infrastructure as service, platform as service and software as service, which can meet the needs of enterprises in different stages, and provide a new development model for modern industry. This paper combines cloud computing technology with traditional equipment maintenance system, and puts forward the cloud platform of equipment maintenance system based on Hadoop, the distributed file system of equipment maintenance, the distributed computing framework of device maintenance, and the resource layer of device maintenance, respectively. The equipment maintenance cloud platform system is discussed in detail in three layers: the equipment maintenance service layer and the equipment maintenance application layer. In this paper, the fault trend prediction module of equipment maintenance service layer is studied, and the support vector regression algorithm is used to predict the fault trend. At the same time, the different effects of parameters Cf8 and 胃 on the performance of support vector regression machine are analyzed. The parameters of support vector regression machine are optimized by particle swarm optimization (PSO). A set of standard data sets in UCI database is used to optimize the experiment. In practical application, the data scale is gradually becoming huge, and the time required for traditional support vector regression machines is increasing dramatically. To solve this problem, a distributed support vector regression algorithm based on Hadoop is proposed. The experimental results show that the prediction performance of distributed support vector regression machine based on Hadoop is basically equal to that of traditional support vector regression machine, and the computing time is greatly saved. At the same time, the influence of increasing the number of Map tasks on time consumption is analyzed under the condition of keeping the sample data unchanged, and it is concluded that increasing the number of Map tasks in a certain range will reduce the time consumption. The prediction model of equipment fault trend based on Hadoop distributed support vector regression machine is established. The prediction performance of the model is verified by using the equipment vibration data collected by a coal enterprise. The results show that the distributed support vector regression machine based on Hadoop has the advantages of saving time, high accuracy and good reliability in fault trend prediction, and it can meet the requirements of practical application.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TH17

【参考文献】