故障特征选择与特征信息融合的加权KPCA方法研究

针对旋转机械振动故障特征与故障类别间不完全对应问题,以双跨转子系统12个通道的故障信息集合为研究对象,提出一种基于加权核主成分分析的故障敏感特征集合提取方法。通过对每个通道的振动信号进行时域、频域、时频域的特征提取,得到一种描述双跨转子系统的原始故障特征集合。采用多准则特征选择方法对这种原始故障特征集合进行特征属性筛选,得到一种利于故障分类的敏感特征集合。对这12个通道的敏感特征集合进行信息融合处理,可得到一种多通道信息的融合特征向量,利用加权核主成分分析方法提取出融合特征向量中的核主成分。结果表明,这种核主成分能够显示出故障类别间的较显著差异,和具有较好的敏感特征子集寻优能力。该研究为解决好双跨转子系统的故障数据集的类别划分问题,提供了一种新途径。     

发电机螺栓松动故障特征检测方法及试验研究

提出一种通过提取时域指标特征和运用主成分分析法(PCA)诊断螺栓松动故障的方法,将原始数据进行经验模式分解(EMD)后计算相应IMF的5个无量纲因子;利用主成分分析法(PCA)对数据向量进行降维和残基空间投影处理,计算数据样本的预测误差,在智能供水系统试验台架上开展螺栓松动故障诊断试验,试验结果表明,所建立PCA模型能够较为有效判别螺栓松动故障。     

主成分分析的入侵检测方法研究

直接将入侵检测算法应用在粗糙数据上,其入侵检测分析的效率非常低.为解决该问题,提出了一种基于主成分分析的入侵检测方法.该方法通过提取网络连接中的相关信息,对它进行解码,并将解码的网络连接记录与已知的网络连接记录数据进行比较,发现记录中的变化和连接记录分布的主成分,最后将机器学习方法和主成分分析方法结合实现入侵检测.实验结果表明该方法应用到各种不同KDD99入侵检测数据集中可以有效减少学习时间、降低各种数据集的表示空间,提高入侵检测效率.     

基于小波函数的核主元特征约简方法研究

提出一种基于小波核函数的核主元特征约简方法,核函数是核主元分析的关键,将Mexican hat小波函数引核主元分析中,以增强核主元分析的非线性映射能力.用转子在正常、油膜涡动、不平衡和径向碰摩状态下的实验数据对该方法进行了检验,比较了主元分析、核主元分析与小波核主元分析的效果.结果表明,小波核主元分析方法能有效地区分转子故障模式,更适合于故障诊断中的非线性特征约简.     

基于模糊神经网络与主成分分析的磨削表面粗糙度在线预测

基于声发射和振动信号提出了一种模糊神经网络和主成分分析的表面粗糙度预测方法,以提高磨削过程中工件表面粗糙度识别的准确性。首先,采集磨削程中声发射与振动信号,提取相关时域特征、频域特征和小波包特征参数,利用主成分分析对特征量进行降维优化;然后,构建表面粗糙度模糊神经网络预测模型,将信号特征量与表面粗糙度作为模糊神经网络的输入和输出;最后,对模型进行训练,并对表面粗糙度预测精度进行验证。实验结果表明:通过主成分分析(PCA)方法对声发射和振动信号特征量进行降维得到5个主成分,以此建立的模糊神经网络表面粗糙度预测模型的效果精度可达到91%以上,与局部线性嵌入和多维标度法降维方法相比,PCA方法降维后的特征所含信息更优,预测准确度更高。     

基于互信息的主成分分析用于声场景分类

主成分分析是声场景分类中常用的特征选择方法。针对主成分分析的局限性,提出一种基于互信息的主成分分析方法。这一方法引入类别信息,用不同声场景条件下特征之间的互信息矩阵之和替代传统主成分分析中的协方差矩阵,计算其特征向量与特征值,特征向量表示由原始特征空间向新的主成分空间的转换系数,特征值则用于计算主成分的累计贡献率并判断主成分维数。声场景分类实验结果表明,该方法较之传统主成分分析方法降维效果更好,辅以神经网络分类器,计算得到的分类正确率更高。     

一种改进的主成分分析方法在股市量价关系研究中的应用

对于多维的股票数据而言,利用改进的主成分分析方法进行数据预处理,在降低股票数据维数的同时还可保留原始数据的绝大部分信息和特征.通过对中国证券市场中的一支股票多维信息进行主成分分析并进行综合评价,找出其中一些简单的量价关系.     

一种基于PCA和贝叶斯分类的气动调节阀故障诊断方法

该文提出一种基于主成分分析(principal component analysis,PCA)和贝叶斯分类的故障诊断方法,并将其应用在气动调节阀的故障诊断中。首先,应用DAMADICS平台仿真气动调节阀多种易发生的故障,监测用于进行故障诊断的信号,采集诊断过程所需要的训练数据集和测试数据集,并对数据集进行主成分分析处理,降低其维度,进而获取数据集的主要特征;然后,利用极大似然估计方法求出训练数据集所满足的多元高斯分布的均值和方差,得到每种故障模式下训练数据集分布的概率密度函数;最后,应用测试数据集进行验证,对于测试数据集中的每个测试数据样本,分别计算测试数据样本属于各种故障类型的后验概率,后验概率越大,对应发生故障的可能性就越大。将该方法与支持向量机(support vector machine,SVM)诊断方法和k-近邻(k-nearest neighbor,k-NN)诊断方法进行对比,诊断准确度整体较高,方法可行。     

基于非线性 PCA的微气体传感器阵列信号处理

在线性叠加模型基础上提出了气体传感器对混合气体的非线性叠加模型,并引入了非线性主成分分析(Nonlinear Principal Component Analysis,NLPCA)法对微传感器阵列的信号进行处理。使用该模型对由4个微热板式气体传感器组成的阵列的信号进行了分析,对照基于线性叠加模型的主成分分析法(Principal Component Analysis,PCA)的识别结果,说明该方法能够提高对混合气体识别和量化的准确度。     

基于主成分分析与BP网络的锚杆支护方案优选

巷道特征与锚杆支护之间可以看作是一种非线性映射关系,用一般的数学方法难以表达巷道支护方案与其影响因素之间的非线性映射关系.神经网络已广泛应用于锚杆支护方案优选,并取得较好的效果.基于单一神经网络预测锚杆支护方案存在一些不足,构建了主成分分析与BP网络相结合的巷道锚杆支护方案优选模型.利用主成分分析对神经网络的输入数据进行预处理,使输入数据减少且不相关,加快网络的收敛速度,并且预测精度均在90%以上.研究结果表明:将主成分分析与BP神经网络结合优选巷道的锚杆支护方案,具有很高的预测精度;与单一BP神经网络相比,提高了预测精度．     

高温热凝结构爆破处理方法的探讨

某炼铁厂高炉在生产过程中产生炉瘤,且有厚度逐渐增厚、范围进一步扩大之势,造成高炉有效容积减少、炉况恶化,制约了高炉的生产能力,影响其正常生产。通过近年来对该公司高炉内部积瘤的爆破施工,总结了处理高温炉瘤的多种方法,并进行了对比,提出了同类工程施工可行的施工方案,为类似工程提供了借鉴。     

北方冬季高炉无料钟炉顶的稳定运行改造

本文主要阐述高炉无料钟炉顶设备的运行使用与生产过程的联系,由于东北冬季室外温度较低,怎样保证炉顶设备在低温气候能满足正常高炉生产运行就是重中之重,根据现场设备不足与其他外界因素的制约,结合与对设备掌握的理论基础和自己的实际工作经验,浅谈几点满足冬季运行的工作手段及措施。     

不停炉情况下的高温炉瘤爆破

为解决某钢铁厂高炉内炉瘤堆积问题,通过爆破方案的选择和爆破参数的计算,确定采用高温氧割烧孔的钻孔爆破法和滑轮吊装药包裸露外敷接触爆破法,同时对爆破振动、飞石及爆炸冲击波进行计算分析,提出相应的爆破安全防护措施,最终利用强制通风、高温冷却、药包隔热等多种辅助手段,在不停炉情况下完成了高温炉瘤的爆破拆除。爆破后,通过对炉壁外侧的红外线探测以及人工观测获知,炉瘤清除较为彻底,爆破效果良好。实践表明,滑轮吊装药包裸露外敷接触及钻孔爆破法可有效去除各类炉瘤,大幅提高除瘤效率与施工安全,改善高炉生产能力,可为解决同类工程难题提供参考。     

SIMULATION OF GAS FLOW IN BLAST FURNACE FOR DIFFERENT BURDEN DISTRIBUTION AND COHESIVE ZONE SHAPE

Blast furnace is a very complex physico-chemical process with countercurrent flow of gas and solid. Computation of gas flow through different regions of blast furnace can be considered as the first major step towards numerical simulation of the process. Gas velocity profile in blast furnace depends on the burden distribution and shape of the cohesive zone, which affect the overall thermal efficiency and utilization of reduction potential of the gas (CO/CO₂ratio). Nonuniform or high gas flow near the wall causes overheating of the wall, which leads to high rate of refractory erosion and heat loss. High gas flow through the central region, on the other hand, causes higher CO/CO₂ratio at the center, leading to reduction in chemical and thermal efficiency of the gas. Therefore, achieving an optimum gas flow by controlling the burden distribution and shape of the cohesive zone is very important for minimizing refractory loss, thermal and chemical efficiency of the process. Gas velocity through various zones of blast furnace is computed by solving vectorial Ergun's equation for flow through porous medium considering axi-symmetric condition. To incorporate the shape of blast furnace efficiently, one-dimensional stretching in radial direction or nonorthogonal body-fitted coordinate system is used by the standard (x,y)to (ξ,η)transformation method. This model can be used for optimizing burden distribution and developing gas velocity control systems for blast furnace.     

高炉钛渣对废润滑油的吸附性能

为研究高炉钛渣的高附加值利用,采用静态吸附法研究了高炉钛渣对废润滑油的吸附性能,考察了高炉钛渣投加量、搅拌转速、吸附温度以及吸附时间对吸附效果的影响,并通过正交实验探讨了高炉钛渣吸附废润滑油的工艺优化条件,在此基础上进行了高炉钛渣与活性白土吸附废润滑油的对比实验。实验结果表明:高炉钛渣对废润滑油有着较好的吸附效果,且吸附效果与活性白土相当,可作为吸附剂用于废润滑油的吸附再生。     

基于主元分析的多联式空调系统传感器故障检测和诊断

作为多元数据分析方法之一,主元分析(PCA)被广泛运用于诊断制冷空调系统的传感器故障。本文首先结合热平衡原理以及多联机运行的控制逻辑,筛选系统中常用的18个传感器变量建立多联机(VRF)传感器的故障分析(FDD)模型。然后结合主元分析的算法原理,给出以Q统计量和Q贡献率为检验标准的传感器故障检测与诊断流程。接着用实测数据进行验证工作,引入不同类型和程度的传感器故障,分析得到不同故障条件下的故障检测和诊断特性。结果表明,总体上,主元分析应用于多联机传感器故障检测与诊断过程是可靠的。其具体特征表现为：不同类型的传感器在不同故障类型及程度条件下,故障检测效果差异明显;在小偏差故障条件下,基于主元分析的传感器故障检测方法的故障检测效率较低,并且针对个别传感器而言,其整体故障检测效率偏低。鉴于故障诊断是基于故障检测的结果,因此上述故障检测方法在FDD过程中将起到重要的作用。     

Physics-data-fusion based decoupling model for coupled faults of complex electromechanical systems

Coupled faults are formed by the nonlinear coupling of multiple lower-level faults in complex electromechanical systems (CES). Although fault decoupling plays a crucial role in locating fault cause and isolating fault components, it still faces challenges due to the harsh reality of common mode failure, networked propagation, and a lack of accurate fault mechanism knowledge in the fault coupling process. A novel physics-data-fusion-based decoupling model for coupled faults of CES was proposed using standard meta components, rigorous formulation, and intuitive representation. First, a hierarchical graph representing the static complex decoupling model was defined by composing proposed meta models. Second, the dynamic model parameters inspired by the time-varying fault characteristics were determined using real-time operation data analysis. Then, based on a proposed numerical reasoning formula, the most likely fault cause was determined, which can also identify fault level by level. Finally, the decoupling model was proved to be reasonable and effective with an offshore wind turbine case. As a graphical modelling method, it handles the decoupling process by fusing static physics and dynamic data of coupled faults. While inheriting the benefits of conventional models, it overcomes the limitations of these existing methods for real-time results. Moreover, the proposed method provided a foundation for tracing the root cause of performance fluctuations, fault detection, and isolation of CES.     

不停炉情况下的高温炉瘤爆破

为解决某钢铁厂高炉内炉瘤堆积问题,通过爆破方案的选择和爆破参数的计算,确定采用高温氧割烧孔的钻孔爆破法和滑轮吊装药包裸露外敷接触爆破法,同时对爆破振动、飞石及爆炸冲击波进行计算分析,提出相应的爆破安全防护措施,最终利用强制通风、高温冷却、药包隔热等多种辅助手段,在不停炉情况下完成了高温炉瘤的爆破拆除。爆破后,通过对炉壁外侧的红外线探测以及人工观测获知,炉瘤清除较为彻底,爆破效果良好。实践表明,滑轮吊装药包裸露外敷接触及钻孔爆破法可有效去除各类炉瘤,大幅提高除瘤效率与施工安全,改善高炉生产能力,可为解决同类工程难题提供参考。     

Evolving Nonlinear Time-Series Models of the Hot Metal Silicon Content in the Blast Furnace

Neural networks are versatile tools for nonlinear modeling, but in time-series modeling of complex industrial processes the choice of relevant inputs and time lags can be a major problem. A novel method for the simultaneous detection of relevant inputs and an appropriate structure of the lower part of the networks has been developed by evolving neural networks by a genetic algorithm, where the approximation error and the number of weights are minimized simultaneously by multiobjective optimization. The networks on the Pareto front are considered possible candidate models that are evaluated on an independent test set. In order to consider the problem of drift in the variables, which may cause parsimonious models to perform poorly on the test set, the weights in the upper layer of the networks are recursively estimated by a Kalman filter. The method is illustrated on a data set from ironmaking industry, where time-series models of the hot metal silicon content in a blast furnace are evolved. The technique is demonstrated to synthesize models with a choice of inputs in agreement with findings presented in the literature and process know-how.     

PIROGAS - a new process allowing diversification of energy sources for blast furnaces

The PIROGAS process consists in replacing, in part or in total, the hot blast blown through the blast furnace tuyeres by reducing gas heated to a temperature of about 2000°C. The heating or the overheating and, eventually, the production of the reducing gas, can be achieved directly in the plasma heater. Up to 80% of the coke required for a conventional blast furnace operation can be replaced by substitute fuels. The PIROGAS process makes it possible to optimize the ratio between fossil fuels and electric power consumption to achieve an economic optimum. Trials carried out on an experimental blast furnace have shown that an operation with co'~e consumption as low as 105 kg of coke/ton of hot metal can be easily attained. Furthermore, it appeared during these trials that the quality of the hot metal can be instantaneously modified by appropriate action on reducing gas temperature and composition. Thus, the control of the process becomes easier and more rapid. Trials carried out on a full size commercial blast furnace have confirmed that the plasma heater in no way disturbs the operation of the blast furnace and, reciprocally, the unavoidable fluctuations in blast furnace operation do not perturb the functioning of the plasma heater. Plasma heaters have a high thermal efficiency (80%) and their electrical behaviour is excellent (power factor 0.92; no harmonics).