【作者】 杨波；

【导师】 付强；

【作者基本信息】 江苏大学 ， 能源动力（专业学位）， 2023， 硕士

【摘要】 重要厂用水泵主要应用于大型核电站的重要厂用水系统当中,其安全性要求很高。而核电站运行安全关乎到诸多与国家发展和安全相关领域,例如清洁能源、水利水电、核能制氢等。其中重要厂用水泵运行工况复杂、故障的可预测性较弱,易导致机组的失修或过渡维修。故对其进行智能监测与故障诊断系统设计是在大力发展核电,增加装机容量的同时也迫切需要解决的问题。针对此问题,本文研究重点是从理论及试验的角度出发,分析了泵体水力、电机电气以及轴承机械故障的信号特征理论规律。构建了实际重要厂用水泵水力故障数据库,并利用LabVIEW完成数据采集、信号处理特征提取以及故障分类算法,最后将其应用到重要厂用水泵状态监测与智能诊断系统中,为核电站主要用泵安全运行提供技术及参考。主要内容如下:1、对重要厂用水泵系统建模;通过CFD模拟计算不同程度叶片折断和口环磨损两种故障得到外特性特征规律。通过磁动势分析计算得到电流信号的频域中转子断条、气隙偏心和匝间短路电机故障特征。利用重要厂用水泵泵用轴承的相关参数,通过公式计算得出内外圈相对旋转频率、保持架、内圈和滚动体频率故障特征、为后续研究提供基础。2、搭建重要厂用水泵试验台;主要包括了试验台架设计以及采集系统的软硬件设备。从多信号融合的角度入手,针对于不同程度的叶片折断、口环磨损故障工况,利用上位机LabVIEW软件搭建采集程序,通过流量计、压力变送器、加速度传感器等硬件设备采集压力、流量、转速和振动等信号并进行分析。3、对重要厂用水泵振动信号进行预处理以及特征提取;以解决重要厂用水泵口环磨损故障原始信号中存在的噪声干扰问题为目的,采用了一种基于集合经验模态分解和小波包阈值的降噪方法,能够有效地提炼出故障信息,提高了振动信号的纯净性。为提高故障的诊断精度,在信号预处理的基础上提取了时域、频域常见的经验指标,提出了小波包特征提取方法以及CEEMD特征提取方法,降低特征维度。4、建立重要厂用水泵智能诊断模型;主要是对上述研究成果中提取到的四种特性进行比较分析,提出了BP神经网络、GA遗传算法优化支持向量机以及PSO遗传算法优化支持向量机三种分类模型。设置学习率、核函数、训练样本等参数对比后分类效果,进一步提升了重要厂用水泵的故障识别准确率。5、搭建重要厂用水泵状态监测与智能诊断系统;综合上述研究方法,利用上位机软件LabVIEW搭建良好的人机交互界面,使用MATLAB与Python实现高精度算法。系统中主要包括对于重要厂用水泵的状态监测与智能诊断功能,主要可以实现多频微弱信号的监测、故障特征的综合处理、常见SEC泵的水力、机械故障诊断等功能。更多还原

【Abstract】 The necessary service water pump is mainly used in the necessary service water system of large nuclear power plants,and its safety requirements are very high.The safety of nuclear power plant operation is related to many fields related to national development and security,such as clean energy,water conservancy and hydropower,nuclear hydrogen production,etc.The operating conditions of necessary service water pumps are complex,and the predictability of faults is weak,which can easily lead to the unit being out of repair or transitioning to maintenance.Therefore,designing an intelligent monitoring and fault diagnosis system for nuclear power is an urgent issue that needs to be addressed while vigorously developing nuclear power and increasing installed capacity.In response to this issue,the focus of this article is to analyze the signal characteristic theoretical laws of pump hydraulic,motor electrical,and bearing mechanical faults from the perspectives of theory and experiment.A database of hydraulic faults in necessary plant water pumps was constructed,and data acquisition,signal processing feature extraction,and fault classification algorithms were completed using LabVIEW.Finally,it was applied to the status monitoring and intelligent diagnosis system of necessary plant water pumps,providing technology and reference for the safe operation of main pumps in nuclear power plants.The main content is as follows:1.Modeling of necessary plant water pump systems;By using CFD simulation to calculate the external characteristic characteristics of different degrees of blade breakage and ring wear faults.The fault characteristics of rotor bar breakage,air gap eccentricity,and inter turn short circuit motor in the frequency domain of the current signal are calculated through magnetomotive force analysis.Using the relevant parameters of bearings used in necessary plant water pumps,the relative rotation frequency of the inner and outer rings,the frequency fault characteristics of the cage,inner ring,and rolling element are calculated through formulas,providing a basis for subsequent research.2.Build an necessary plant water pump test bench;This mainly includes the design of the test bench and the software and hardware equipment of the acquisition system.Starting from the perspective of multi-signal fusion,for different degrees of blade breakage and ring wear fault conditions,a collection program is built using the upper computer LabVIEW software.The pressure,flow,speed,and vibration signals are collected and analyzed through hardware devices such as flow meters,pressure transmitters,and acceleration sensors.3.Pre processing and feature extraction of vibration signals from necessary plant water pumps;A denoising method based on set empirical mode decomposition and wavelet packet threshold was adopted to solve the noise interference problem in the original signal of the wear fault of the necessary plant water pump mouth ring.This method can effectively extract fault information and improve the purity of the vibration signal.To improve the accuracy of fault diagnosis,common empirical indicators in the time and frequency domains were extracted based on signal preprocessing.The wavelet packet feature extraction method and CEEMD feature extraction method were proposed to reduce the feature dimension.4.Establish an intelligent diagnostic model for necessary plant water pumps;The main focus is to compare and analyze the four characteristics extracted from the above research results,and propose three classification models: BP neural network,GA genetic algorithm optimized support vector machine,and PSO genetic algorithm optimized support vector machine.By comparing the classification results with parameters such as learning rate,kernel function,and training samples,the accuracy of fault identification for necessary plant water pumps has been further improved.5.Establish an necessary plant water pump status monitoring and intelligent diagnosis system;Based on the above research methods,a good human-machine interaction interface is built using the upper computer software LabVIEW,and high-precision algorithms are implemented using MATLAB and Python.The system mainly includes status monitoring and intelligent diagnosis functions for necessary plant water pumps,which can achieve monitoring of multi frequency weak signals,comprehensive processing of fault features,hydraulic and mechanical fault diagnosis of common SEC pumps,and other functions.更多还原