基于EMD样本熵和BP神经网络的乳化机故障诊断系统研究

发布时间：2018-04-29 14:40

本文选题：乳化机 + EMD　；参考：《杭州电子科技大学》2017年硕士论文

【摘要】：乳化炸药作为一种新式环保型炸药,具有爆轰猛度强、抗水性良好等特点。同时乳化炸药生产工艺简单、产能大、生产成本低使得它在我国民爆行业得到了广泛应用。乳化机是乳化炸药连续化、自动化生产线的核心设备,它是一种改良的旋转机械设备,且具有高速旋转的特性。要是内部器械运转异常,不仅会破坏生产工序的连续性能,影响产能和质量,严重情况下还会发生机毁人亡的安全事故,造成巨大的经济损失和社会影响。为了能准确地检测出乳化机潜藏的故障,提高维修效率,保证设备安全,本文研制了一套乳化机故障检测和诊断系统,并在实际生产中得到成功应用。本文主要针对转子故障、轴承故障等常见故障类型,在研究各类型的故障机理和发生征兆的基础上,首先提出基于样本熵的振动信号故障特征提取方法。针对样本熵对原始信号获取有限,故障特征区分度不高的缺陷,提出经验模式分解方法(Empirical Mode Decomposition,简称EMD)预处理样本熵的故障特征提取方法。该方法利用EMD先把振动信号分解为若干个固有模态函数(Intrinsic Mode Function,简称IMF),然后选取若干具有代表性的IMF分量,将这些分量的样本熵组成向量作为故障特征。EMD方法能将蕴藏在信号内部的信息挖掘出来,有效克服样本熵对信息获取的局限性。结果表明,EMD结合样本熵的方法不仅能够区分不同类型的故障种类,还能提高了识别系统的容错率。神经网络具备强非线性映射,以及自学习、自组织和自适应的能力。将提取的故障特征作为BP神经网络的输入,通过整理乳化机正常和故障的振动历史数据,分别构造了振动特征参数的正常及故障状态的训练样本,并用训练好的神经网络进行故障类型识别。结果表明,BP神经网络能快速地识别出滚动轴承的故障类型,诊断效果良好。在乳化炸药生产线现有设备的基础上完成硬件配置,基于工业控制软件组态王平台实现对PLC的数据交互,通过VB调用MATLAB神经网络功能实现上位机故障诊断系统的开发。实践结果表明,本文构建的乳化机故障诊断系统能够根据实际数据准确地识别出乳化机的故障类型,诊断准确率高,实际应用效果好。
[Abstract]:Emulsion explosive, as a new type of environmental protection explosive, has the characteristics of strong detonation intensity and good water resistance. At the same time, emulsified explosive has been widely used in our country because of its simple production process, large production capacity and low production cost. Emulsifying machine is the core equipment of continuous and automatic production line of emulsion explosive. It is an improved rotating machine and has the characteristics of high speed rotation. If the operation of internal instruments is abnormal, it will not only destroy the continuous energy of production process, but also affect the production capacity and quality. In serious cases, the safety accident of machine destruction and death will occur, resulting in huge economic loss and social impact. In order to accurately detect the hidden faults of emulsifying machine, improve the maintenance efficiency and ensure the safety of the equipment, a fault detection and diagnosis system for emulsifying machine has been developed in this paper, and has been successfully applied in practical production. This paper mainly aims at the common fault types such as rotor fault bearing fault and so on. On the basis of studying the fault mechanism and occurrence symptom of each type a method of vibration signal fault feature extraction based on sample entropy is proposed. Aiming at the defects of limited sample entropy acquisition and low fault feature differentiation, an empirical Mode decomposition (EMD) preprocessing sample entropy method for fault feature extraction is proposed. In this method, the vibration signal is first decomposed into intrinsic Mode functions by EMD, and then some representative IMF components are selected. The sample entropy component vector of these components can be used as fault feature. EMD method can mine the information hidden inside the signal and overcome the limitation of sample entropy to information acquisition. The results show that EMD combined with sample entropy can not only distinguish different types of faults, but also improve the fault tolerance of the identification system. Neural networks have strong nonlinear mapping, as well as self-learning, self-organization and adaptive capabilities. Taking the extracted fault features as the input of BP neural network, the normal and fault training samples of the vibration characteristic parameters are constructed by sorting out the history data of the normal and fault vibration of the emulsifier. The trained neural network is used to identify the fault types. The results show that BP neural network can quickly identify the fault types of rolling bearings, and the diagnosis effect is good. The hardware configuration is completed on the basis of the existing equipment in the emulsion explosive production line, the data exchange to PLC is realized based on the industrial control software Kingview platform, and the development of the upper computer fault diagnosis system is realized by calling the function of MATLAB neural network in VB. The practical results show that the emulsifying machine fault diagnosis system constructed in this paper can accurately identify the fault types of the emulsifier according to the actual data. The diagnosis accuracy is high and the practical application effect is good.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TQ560.5;TP183;TP277

【参考文献】