基于神经网络的风机故障诊断研究
发布时间：2018-07-26 11:46
【摘要】：风机在工业生产中发挥着重要的作用，当风机发生故障时，不仅对整个生产线产生直接影响，而且会造成重大的经济损失甚至是机毁人亡的事故。为保证设备的安全运行,降低机组维修费用和提高设备利用率，设计出一种自动获取知识且能进行高速推理的故障诊断方法，已经成为风机故障诊断研究的一个主要方向。 本论文针对某炼钢厂风机的故障诊断与状态监测进行研究，采用PDM2000数据采集分析仪对故障风机进行振动信号采集，获得特征频率，根据设备振动诊断技术的频谱分析方法，分析讨论风机故障的故障征兆,得到风机故障的诊断结果。同时又采用BP神经网络分析方法，对风机的故障做进一步的诊断分析。 本文根据BP神经网络的结构形式及算法，选用三种方法对BP神经网络的算法进行改进；并通过实测数据运算及三种改进算法的相互比较，从而选出运算速度比较快、判断比较准确的Levenberg-Marquardt算法对所建立的BP神经网络进行训练分析。将采集的现场风机的特征数据，通过Matlab软件进行训练；并通过已训练完成的BP神经网络对其进行测试，从而判断得出风机目前也存在转子不平衡、转子碰摩及轻微转子不对中等故障，，其诊断结果与现场实测分析结果相吻合。 本文通过实测的风机振动数据分析结果与理论计算结果进行比较分析,证明本文提出的采用BP神经网络改进算法对风机故障进行诊断具有一定的实用性和可行性。
[Abstract]:Fan plays an important role in industrial production. When the fan breaks down, it will not only have a direct impact on the whole production line, but also cause great economic losses and even fatal accidents. In order to ensure the safe operation of the equipment, reduce the maintenance cost of the unit and improve the utilization rate of the equipment, a fault diagnosis method which can automatically acquire knowledge and carry out high-speed reasoning has become a main research direction of fan fault diagnosis. In this paper, the fault diagnosis and condition monitoring of fan in a steelmaking plant is studied. The vibration signal of the fan is collected by PDM2000 data acquisition analyzer, and the characteristic frequency is obtained. According to the frequency spectrum analysis method of the equipment vibration diagnosis technology, the frequency spectrum of the fault fan is obtained. The fault symptom of fan fault is analyzed and the diagnosis result of fan fault is obtained. At the same time, BP neural network analysis method is used to diagnose fan fault further. According to the structure and algorithm of BP neural network, three methods are selected to improve the algorithm of BP neural network. A more accurate Levenberg-Marquardt algorithm is used to train and analyze the BP neural network. The characteristic data of the field fan are trained by Matlab software, and tested by BP neural network which has been trained, and it is judged that the fan also has rotor unbalance at present. The results of diagnosis are in good agreement with the measured results. In this paper, the analysis results of the measured fan vibration data and the theoretical calculation results are compared and analyzed. It is proved that the improved BP neural network algorithm proposed in this paper is practical and feasible for fan fault diagnosis.
【学位授予单位】：辽宁科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TH165.3
本文编号：2145908
[Abstract]:Fan plays an important role in industrial production. When the fan breaks down, it will not only have a direct impact on the whole production line, but also cause great economic losses and even fatal accidents. In order to ensure the safe operation of the equipment, reduce the maintenance cost of the unit and improve the utilization rate of the equipment, a fault diagnosis method which can automatically acquire knowledge and carry out high-speed reasoning has become a main research direction of fan fault diagnosis. In this paper, the fault diagnosis and condition monitoring of fan in a steelmaking plant is studied. The vibration signal of the fan is collected by PDM2000 data acquisition analyzer, and the characteristic frequency is obtained. According to the frequency spectrum analysis method of the equipment vibration diagnosis technology, the frequency spectrum of the fault fan is obtained. The fault symptom of fan fault is analyzed and the diagnosis result of fan fault is obtained. At the same time, BP neural network analysis method is used to diagnose fan fault further. According to the structure and algorithm of BP neural network, three methods are selected to improve the algorithm of BP neural network. A more accurate Levenberg-Marquardt algorithm is used to train and analyze the BP neural network. The characteristic data of the field fan are trained by Matlab software, and tested by BP neural network which has been trained, and it is judged that the fan also has rotor unbalance at present. The results of diagnosis are in good agreement with the measured results. In this paper, the analysis results of the measured fan vibration data and the theoretical calculation results are compared and analyzed. It is proved that the improved BP neural network algorithm proposed in this paper is practical and feasible for fan fault diagnosis.
【学位授予单位】：辽宁科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TH165.3
【参考文献】
相关期刊论文 前4条
1 郑君;张冬泉;;故障诊断技术[J];电气时代;2008年05期
2 黄永东;;转子不平衡现象的分析[J];发电设备;2009年03期
3 刘占生;赵广;龙鑫;;转子系统联轴器不对中研究综述[J];汽轮机技术;2007年05期
4 周政;;BP神经网络的发展现状综述[J];山西电子技术;2008年02期
本文编号：2145908
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2145908.html
