【作者】 臧红岩；

【导师】 马凤英；

【作者基本信息】 山东轻工业学院 ， 检测技术与自动化装置， 2011， 硕士

【摘要】 随着机械设备的日趋复杂,其日常生产中的可靠性和安全性及维护问题已经引起人们的极大关注。能够建立准确地定位和及时地诊断故障诊断系统是解决这一问题的关键。矿井通风机是煤矿企业的重要供风设备,其故障是导致瓦斯爆炸的主要原因之一。因此,对矿井通风机实施故障检测与诊断具有重大意义。本文以粗糙集和人工神经网络等理论为基础,重点研究了故障特征约简和BP神经网络在风机故障诊断中的应用,以基于粗糙集--神经网络的智能诊断方案,实现对矿井通风机常见机械故障的有效诊断。本文主要从以下几个方面作了探索性尝试:(1)根据粗糙集和神经网络各自的优缺点,选择矿井风机作为基于两者相结合的故障诊断模型的研究对象。在粗糙集理论基础上,讨论了数据的离散化问题、样本数据的约简问题。(2)对BP神经网络理论及结构和算法做了相应分析研究,了解了神经网络集成思想和模式,以及故障诊断中集成神经网络的应用,并利用MATLAB的神经网络工具箱进行数据训练及仿真。(3)选取UCI数据集中提供的风机故障样本作为本课题研究中仿真的原始数据样本,应用粗糙集的约简方法对大样本进行属性约简,结合BP神经网络完成矿井风机的故障诊断。(4)通过神经网络诊断结果和粗糙集--神经网络的诊断结果的对比,在简化网络结构的和提高训练速度方面做了分析,通过实例仿真说明约简对于大样本故障诊断是可行的。最后,证明了利用粗集作为神经网络诊断的数据预处理在风机故障诊断中的可行性及优势。更多还原

【Abstract】 With the increasingly complex of mechanical equipment, the reliability and safety and maintenance problems of daily production have aroused people’s attention. In order to solve this problem,build timely and accurate positioning and diagnosis fault diagnosis system is the key. Coal mine ventilator is important for enterprise’s ventilation equipment, its fault is one of the main reasons led to the gas explosion. Therefore, it’s a great significance to implementation of fault detection and diagnosis for mine ventilator. Based on rough sets and artificial neural network theory, focus on the fault feature reduction and the BP neural network fault diagnosis in the application of fan , proposed a scheme of intelligent diagnosis of the mine ventilator based on the rough set - neural network, in order to realize common mechanical fault diagnosis effectively.This article try several aspects exploratory as follows:(1) Based on the advantages and disadvantages of rough sets and neural network ,choose mine fan as research object for this fault diagnosis model. Discussed the discretization and reduction problem, illustrates its related problems such as classification ability and decision accuracy .(2) Study and analyzed the structures and algorithms of the BP neural network theory and Integrated neural network. Achieves the training and simulation using the neural network toolbox of MATLAB .(3) In this topic, selected the dataset UCI sample as fan fault sample, used the rough sets reductied the sample, and combined with BP neural network finished the fault diagnosis of mine fan.(4) Compared the results between the two methods, which the neural network diagnosis and rough set neural network diagnosis, an example show the diagnosis model is feasible for Large sample data.Finally, It is proved that neural network using rough sets is feasibility and advantage of fault diagnosis.更多还原