【作者】 侯军虎；

【导师】 王松岭；

【作者基本信息】 华北电力大学（河北） ， 热能工程， 2004， 博士

【摘要】 通风机属于通用机械范畴,在国民经济的各部门应用十分广泛。尤其在火电厂中,通风机是烟风系统的动力源,其运行状况直接关系到电厂的安全、经济运行,因此实施对通风机状态监测与故障诊断具有重大意义。目前该课题研究是以机械特性为研究主体、以对振动信号的故障诊断为主要内容而展开的。而综合空气动力性能、机械特性和非稳态流动特性的通风机状态监测与故障诊断技术在国内外还没有得到系统的研究。基于此,本文采用实验研究与理论分析结合的方法,在实现对电站通风机的空气动力性能监测、非稳态流动特性分析和机械故障特征提取的基础上,采用统计模式识别的方法对多指标特征进行有效融合与压缩,并与神经网络结合提出了单网络和子网络的诊断模型,实现了对多特性故障、故障的不同严重程度和耦合故障的有效诊断,开发了基于多参数的离心式通风机状态监测与故障诊断系统。1.基于参数映射的风机流量全程测量的实验研究。基于通风机无因次性能曲线反映出的性能参数间稳定的映射关系,提出了四种曲面逼近方法并对其逼近性能进行了比较说明:RBF网络具备更优越的逼近特性。文中选用RBF网络逼近风机性能曲线不同区域的非线性映射关系,进而完成了基于参数映射的流量测量模型的推导。并在实验室对该模型的应用进行了实验研究,对于风机相似定律适用范围与该模型的内在关系等与之相关的问题一并进行了实验研究,得到了该模型在相似定律范围内良好的测量精度。从而实现了基于参数映射的通风机流量的无节流准确在线监测。2.电站离心风机旋转失速的实验研究及特性分析。结合离心风机叶栅的流场状况分析了两类旋转失速的形成机理,在4-73No8D离心风机上进行了旋转失速实验研究,构造了非稳态流动实验测试系统,进行了旋转失速静态特性与渐进过程特性实验。通过实验研究,在4-73风机上发现了压力侧失速现象,并对失速的幅频域特性以及对性能参数的影响等问题进行了研究。基于失速动态特性分析的需要,采用谐波小波进行时频分析,提出了改进的非正交谐波小波,并给出了其快速算法。分别采用正交和改进的非正交谐波小波变换对于弱失速特性、渐进失速过程时频特性等进行了分析研究,给出了相关结论。研究结果也显示出了谐波小波变换在风机旋转失速特性分析方面的优越性。3.通风机故障的实验模拟和故障诊断的多指标特征生成。在离心风机实验台上对于典型的机械和非稳态流动故障进行实验模拟,构建了风机故障实验的测试系统,对机械、空气动力特性参数进行了测量、变换和处理,为故障诊断提供了逼近于现场运行风机的故障样本集。在此基础上,分别就机械振动信号的谐波特征、能量特征、碰摩故障奇异性定性和定量特征生成,以及非稳态流动故障识别中动态压力信号频率与性能参数联合特征生成等问题采用小波变换等方法进行了研究,形成<WP=5>了风机故障诊断的多参数多指标特征,并给出了特征的预处理方法。4.基于统计模式识别方法的风机故障诊断特征压缩和神经网络分类。采用特征选择方法进行了特征压缩与融合,选择兼备均值差别和方差差别的Bhattacharyya距离作为准则函数,采用综合类别对之间判别能量比较的特征选择算法,推导出了B-特征选择的特征压缩方法,并与具备优良边界特性的单隐层前向网络构成B-单网络诊断模型,以此作为风机单故障的诊断模型。对耦合故障诊断问题提出了B-子网络诊断模型,并应用实验样本验证了B-特征选择算法和B-子网络诊断模型的优越性。5.基于多参数的风机状态监测与故障诊断系统(FMMDS)的开发。开发了FMMDS的软硬件系统,该系统具备多特性多参数监测,多指标特征生成与压缩、多功能信号分析、管网阻力状态评价、风机运行经济性分析、单故障与耦合故障的诊断,以及高速同步采样与DCS通讯等多种功能。在FMMDS的工程实践中,针对某电厂的200MW机组送引风机,开发了基于DCS的通风机状态监测与分析系统FMAS1.0。更多还原

【Abstract】 The ventilating fan is one type of large rotating machinery in power plant and also is the power supply of the air/gas system, and the operating condition of which has obvious influence to security and economic of units. Therefore it is necessary to take effective monitoring and diagnosis on ventilating fan. Nowadays the study of monitoring and fault diagnosis of fan is mainly focus on vibration signal with respect to mechanical characteristic based on Fourier analysis. Till now the study of condition monitoring and fault diagnosis based on multi-parameter and multi-characteristic is attracting much more attention. But for a ventilating fan, multi-property condition monitoring and fault diagnosis integrating aerodynamic performance, mechanical characteristic and unstable flow property isn’t systematically studied yet.In the paper the experimental study means is adopted. Aerodynamic performance monitoring is realized and characteristic analysis of unstable flow and feature extraction of mechanical trouble are carried through in this paper. Based on which multi-parameter condition monitoring and fault diagnosis of fan is put forward and the high dimensional features of multi-index and multi-parameter are reduced and fused by the feature selection means of statistical pattern recognition, and different neural network combined with which can deduce the single network and subnetwork fault diagnosis model. The proposed model accurately diagnoses the multi-property troubles, different severity troubles and coupling troubles using experimental samples. And based upon which the multiparameter condition monitoring and fault diagnosis system is developed in the paper.1. Experimental investigation on flow rate measurement of fan based on parameter mapping. It has been proved by our former study that the steady mapping law between performance parameters can be reflected by the dimensionless performance curves of a fan. In the paper four different approximating methods are put forward and the approximating performance of which is compared. It is proved that RBF network has better approximating performance. Therefore the RBF network is used for approaching the different nonlinear parameter mapping law and based on which the parameter mapping flow measurement model is derived. Aiming at the application of parameter mapping flow rate measurement model of fan the experiment is taken through on a 4-73No8D fan in the lab. And the related problem such as the similarity law application range of fan is also experimentally studied. It is verified that in the range of similarity law the measurement based on the model obtains high accuracy and the application range of similarity law of fan on the condition of speed change is the main influencing factor of model.2. Experimental investigation and characteristic analyze for rotating stall of centrifugal fan. Based on analyze for the flow field conditions of blade row of centrifugal fan the formation mechanism of two kinds of rotating stall (s-stall and p-stall) is explained. The unstable flow test system is developed and experimental investigation is carried through on a 4-73No8D centrifugal fan. The rotating pressure stall (p-stall) is found on a 4-73 fan for the first time. In the experiment the stall property in amplitude and frequency domain and the influence to performance parameter of stall are studied. In order to explaining the dynamic property of rotating stall Harmonic Wavelet Transform (HWT) is adopted for time-frequency characteristic analysis. The reformed HWT that has the adjustable time resolution is derived <WP=7>and its fast algorithm is also given in the paper. The orthogonal HWT and non-orthogonal HWT are respectively applied to analyze the feature of weak stall and progressive stall procedure and so on. The conclusion is given and the study results are also proved the advantage of HWT in analyzing the characteristic of rotating stall.3. The trouble simulating of fan and multi-index feature generation of fault diagnosis. Some classical mechanical troubles and unstab更多还原