Expert System for the Design of Single Speed Gear Boxes

Based on the characteristic peculiarities of mechanical design expert systems (MDES), the design process pf gear box and its components is introduced and the gear box design expert systems (GBES)is established. GBES employs the methods of knowledge representation to indicate the knowledge-unit-rule-process, table-vector-process. By taking the advantage of knowledge unit's indicator, it can make the units of knowledge base to to combine to form a whole in the feature of trees and nets so that it can give deduction conveniently. The knowledge base of GBES is organized in hierarchy, which provides the efficient managerial systems of knowledge base. It makes the knowledge base convenient greatly for establishing and using. The assistant modules of GBES are written in FORTRAN and the part of expert systems is written in LISP. It explains the I/O among each module and the forms of independent application. The GBES systems have been put into preliminary, use in practice.

Study on Rolling Mill Gear Box and Design of Multichannel Fault Diagnosis

Gear box places an important role rolling mill.Its reliability decides the machine operation.Due to the important role,if the key machine is broken because of gear box′s malfunction,the whole production devices will be influeued.Therefore,it′s very important to monitor the gear box online.Good monitoring system can help companies to better process fault diagnosis.The design sets up a monitoring system with Enwatch polled mode on-line acquisition module and Odyssey software.By calculating the data,the problem of the monitoring system is find,the plans to collect signal is made,the problems of monitoring gear box′s multichannel vibration are solved and the malfunctions initially are estimated according to the signal,which has theoretical basis and practical meanings.

Wavelets applied to early fault analysis of hoist gear box

Using wavelets, the vibration signal of a certain mine hoist gear box was analyzed. By multiple comparison analysis, the rational wavelet basis function was determined. Fault characteristic frequencies of hoist gear box were identified. The research indicates that the hoist＇s fault information is non-stationary, and non-stationary signal is clearly extracted by using db20 wavelet as basis function. The db20 wavelet is the proper wavelet base for vibration signal analysis of the hoist gear box.

Gearbox Scheme in High Temperature Reactor Helium Gas Turbine System

Helium Turbine is used in High Temperature Reactor Helium Gas Turbine (HTR-GT) system, by which the direct helium circulation between the reactor and turbine generator system will come true. Between helium turbine and generator, there is gearbox device which reduces the turbine rotation speed to normal speed required by the generator. Three optional gearbox schemes are discussed. The first is single reduction cylindrical gearbox, which consists of one high speed gear and one low speed gear. Its advantage is simple structure, easy to manufacture, and high reliability, while its disadvantage is large volume and misalignment of input and output axle. The second is planetary gear mechanism with static planet carrier. The third is planetary gear mechanism with static internal gear. The latter two gearbox devices have similar structure. Their advantage is small volume and high reduction gear ratio, while disadvantage are complicated structure, many gears, low reliability and low mechanical efficiency.

GEAR BOX CONCEPT DESIGN EXPERT SYSTEM OF NC MILLER BASED ON OBJECT─ORIENTED TECHNOLOGY

The method and basic structure of object─oriented knowledge representation are presented in detail and the hierarchical structurc of object description is explained Based on object─oriented technology.the gear box concept design expert cistem of NC miller (GBCDLS) has been developed and applied to practical use.Moreover. the design principlc and techniques of GBCDES are discussed and analyzed. and the main functions of GBCDES are described. From the results of applications.it can be seen that GBCDES is effective and successful.

Analysis of Electromagnetic Performance of Modulated Coaxial Magnetic Gears Used in Semi-Direct Drive Wind Turbines

Wind turbine is a key device to realize the utilization of wind energy,and it has been highly valued by all countries.But the mechanical gear transmission of the existing wind power device has the disadvantages of high vibration and noise,high failure rate,and short service time.Magneticfield modulation electromagnetic gear transmission is a new non-contact transmission method.However,the conventional modulation magnetic gear has low torque density and torque defects with largefluctuations.In order to overcome the gear transmis-sion problems of the existing semi-direct drive wind power generation machinery and improve the electromag-netic performance of the traditional magnetic gear transmission,this paper proposes a new transmission scheme of a non-contact semi-direct drive wind generator with a surface mount Halbach array modulated mag-netic gear method,and considers the electromagnetic properties of the semi-direct drive modulation magnetic gear of the wind turbine.Thefinite element software is used to construct the model of the surface-mounted Halbach array magnetic gear and the conventional gear,analyzed the distribution of magneticfield lines of the two magnetic gears,calculated the air gap magneticflux density of the inner and outer air gap,and obtained the main harmonics of the inner and outer air gap magnetic density;calculated the static torque and steady-state operating torque of the inner and outer rotors in the model,compared the air gapflux density,harmonics and torque of the magnetic gears.The simulation results show that the magneticfield modulation type mag-netic gear of the surface mount Halbach array magnetic gear method improves the magnetic induction wave-form of the inner and outer air gap,reduces the pulse torquefluctuation,and has a 60%higher static torque.Applying it to semi-direct drive wind power generation equipment not only overcomes the shortcomings of mechanical gears,but also has higher electromagnetic performance.Therefore,the surface-mounted Halbach array modulated magnetic gear can be used to replace the mechanical gearbox in the semi-direct drive wind power generation equipment.

基于格拉姆角场与深度卷积生成对抗网络的行星齿轮箱故障诊断

针对行星齿轮箱故障诊断中样本分布不均衡所引起的模型泛化能力差及诊断精度低等问题,采用格拉姆角场图像编码技术和深度卷积生成对抗网络相结合进行数据增强,融合AlexNet卷积神经网络进行故障诊断。将采集到的一维振动信号转化为格拉姆角场图,按比例划分训练集与测试集,将训练集样本与随机向量输入到深度卷积生成对抗网络模型中,交替训练生成器与判别器,达到纳什平衡,生成与原始样本类似的生成样本,从而实现故障样本的增广。用原始样本与生成的增广样本训练卷积神经网络分类模型,完成行星齿轮箱的故障识别。实验结果表明,所提方法能够有效提升样本不均衡条件下的行星齿轮箱故障诊断精度,使之达到99.15%,且能使收敛速度更快。

基于MTF-Swin Transformer的风机齿轮箱故障诊断

针对风机齿轮箱实际工况复杂多变及含有强噪声,传统故障诊断方法对风机齿轮箱故障诊断识别准确率较低的问题,文章提出了MTF-Swin Transformer风机齿轮箱故障诊断模型。首先,采用马尔科夫变迁场(MTF)图形编码方法将原始一维振动时序信号转化为具有关联时间信息的二维特征图谱;然后,将特征图谱作为Swin Transformer模型的输入,基于自注意力机制进行自动特征提取;最后,实现对不同故障类型的分类。仿真结果表明,该方法对齿轮箱故障诊断准确率达到了99.48%,证明了该方法的有效性和优越性。

基于HMFDE和t-SNE的旋转机械故障诊断方法

针对旋转机械的故障特征提取较难,以及单一类型的特征无法全面反映故障特性的问题,提出了一种基于混合多尺度波动散布熵(HMFDE)、t分布-随机邻域嵌入(t-SNE)和郊狼优化算法(COA)优化极限学习机(ELM)的旋转机械故障诊断方法。首先,采用特征加权提出了混合多尺度波动散布熵方法,并将其用于提取旋转机械振动信号的故障特征;随后,采用t-SNE方法对混合故障特征进行了特征降维,挑选出了最能够反映故障特性的特征子集,构建了敏感特征样本;最后,采用郊狼优化算法对极限学习机的输入权重和隐含层阈值进行了优化,完成了旋转机械的故障识别和分类;以齿轮箱和滚动轴承故障数据集为对象,对基于HMFDE、t-SNE和COA-ELM的故障诊断方法进行了实验,验证了方法的有效性。研究结果表明:采用HMFDE-t-SNE-CAO-ELM故障诊断方法可以取得100%的故障识别准确率,该方法能够有效地诊断旋转机械的不同故障类型和损伤;相较于基于单一类型特征的故障诊断方法,其准确率分别可以提高0.68%、22.42%、29.18%(齿轮箱)和1.43%、8.23%、23.67%(滚动轴承),虽然牺牲了一定的计算效率,但准确率得到了明显的提高;相较于其他常规故障分类器,COA-ELM的故障识别准确率具有明显的优势。

IMIBSE与ISOMAP在旋转机械故障诊断中的应用

针对基本熵的区域划分标准不理想,导致无法有效测量振动信号的复杂度,使故障诊断的准确率不佳这一问题,提出了一种基于改进多尺度改进基本熵(IMIBSE)、等距特征映射(ISOMAP)和随机森林(RF)的旋转机械故障诊断方法。首先,采用基于方差的区域划分准则对基本熵进行了改进,结合改进的粗粒化处理,提出了IMIBSE,并将其用于提取故障特征;随后,利用ISOMAP对原始故障特征进行了特征降维,选择了对分类贡献最大的一组特征作为故障敏感特征;最后,基于RF建立了多故障分类器,将故障敏感特征输入至RF模型进行了训练和测试,实现了旋转机械的故障识别,利用齿轮箱和离心泵两种故障数据集将IMIBSE方法与复合多尺度基本熵、多尺度改进基本熵、多尺度基本熵进行了比较和分析。研究结果表明:IMIBSE不仅具有最佳的可视化效果,而且取得的识别准确率最高,二者均达到了100%,而二者的平均分类准确率分别为100%和99.8%;相较于其他故障诊断方法,IMIBSE方法的准确率更高,而且适用于小样本的故障识别问题。

基于维纳滤波的风电机组齿轮箱轴承故障诊断

存在轴承故障时,风电机组齿轮箱振动信号的幅度与频率都会发生明显改变,通过信号变化行为的测量能够判定出具体的轴承故障等级。为实现对轴承故障的准确判定,设计基于维纳滤波的风电机组齿轮箱轴承故障等级判定方法。通过获取轴承振动波的IMF分量,并完善维纳滤波算法,在此基础上,处理故障数据,实现基于维纳滤波的轴承故障特征提取。定义轴承故障特征的伪四阶矩,通过求解特征角度的方式,确定故障等级基准的判定范围,完成基于维纳滤波的风电机组齿轮箱轴承故障等级判定方法的设计。实验结果表明,利用上述方法,能够根据幅度和频率的变化行为,判定风电机组齿轮箱轴承的故障等级,符合实际应用需求。

基于Transformer和卷积神经网络的齿轮故障诊断方法

针对部分齿轮的运行环境复杂,导致采集的样本数据不够的问题,提出了一种基于Transformer和卷积神经网络(CNN)的迁移学习齿轮故障诊断方法。首先,采用高斯滤波对原始振动信号进行了预处理,使信号变得平滑,降低了噪声信号的干扰;再将信号处理成带有位置信息的补丁序列以作为Transformer的输入,并增强了Transformer特征提取的能力,提高了诊断精度;然后,将信号输入到CNN继续提取特征信息,在模型中添加了一个残差块以防止网络退化;接着,划分了实验室采集的齿轮数据集和东南大学齿轮箱数据集的源域和目标域,采用了源域数据预训练模型,选择了每种类型的齿轮各100个样本为目标域;最后,以不同数据集为源域共进行了4组10次重复实验,测试了模型的准确率。研究结果表明:以不同数据集为源域的4组10次迁移实验的齿轮故障诊断准确率较高,均在90%以上,最高准确率可达100%;与其他不含Transformer的卷积神经网络、多尺度卷积神经网络和二维卷积神经网络相比,Transformer-CNN的齿轮故障诊断平均准确率更高,其平均准确率可达到99.64%。因此,基于Transformer-CNN的迁移学习方法能在小样本下诊断齿轮的故障。

基于CMMFDE与多传感器信息融合的旋转机械故障诊断研究

采用单一传感器采集的振动信号难以准确描述旋转机械动态特性,导致提取的故障特征无法准确辨识旋转机械故障。针对这一缺陷,提出了一种基于复合多元多尺度波动散布熵(CMMFDE)、多传感器信息融合和哈里斯鹰算法优化极限学习机(HHO-ELM)的旋转机械故障诊断方法。首先,引入复合多元粗粒化处理,提出了CMMFDE方法,避免了传统单变量分析方法只能处理单一通道振动信号而导致特征的表征性能不足的缺陷,增强了故障特征的表征性能;随后,利用布置在旋转机械不同部位的传感器收集了多种类型的信号,组成混合多通道信号,并进行了CMMFDE分析,构建了故障特征;最后,采用HHO对极限学习机的参数进行了自适应优化,并对特征样本进行了训练和测试,完成了旋转机械的故障识别工作;利用齿轮箱、离心泵两种典型的旋转机械数据集进行了实验分析。研究结果表明:该方法对多个通道的信号进行分析时,所获得的准确率达到了100%和98%,优于对单个通道信号进行分析时获得的准确率,同时CMMFDE方法的准确率和特征提取时间均优于精细复合多元多尺度熵(RCMMSE)、精细复合多元多尺度模糊熵(RCMMFE)、精细复合多元多尺度排列熵(RCMMPE)、多元多尺度波动散布熵(MMFDE)。

基于EMD分解和Levy-SSA-BP神经网络的齿轮故障诊断

为解决齿轮磨损早期故障诊断问题,提出了一种基于经验模态分解(Empirical Mode Decomposition,EMD)和算法优化反向传播(Back Propagation,BP)神经网络的故障诊断方法。首先,将声发射信号进行经验模态分解,得到本征模函数(Intrinsic Mode Function,IMF);其次,计算各IMF分量与原始信号的相关系数,并对各个分量进行特征提取,构成特征矩阵;最后,将特征矩阵放入经过Levy飞行和麻雀搜索算法优化后的BP神经网络中进行识别。对比BP神经网络和麻雀搜索算法优化后的神经网络,本文提出的算法准确率更高,且对轻微磨损故障的识别能力更好,可以用于早期齿轮故障诊断。

基于SCADA数据协同的风力机故障检测与控制

为了在不停机的前提下,安全可靠地对齿轮箱进行故障检测,并实现风力发电机在齿轮箱故障工况下的容错控制运行,提出一种基于标准数据采集与监视控制(SCADA)数据协同的风力机故障检测与控制方案。介绍了风力发电机的系统模型与SCADA解决方案框架,通过回归建模、异常分析和集成学习对风力发电机系统进行故障检测并获得健康指标;借助模糊逻辑控制对风力发电机的输出功率进行降额控制,从而实现风力发电机在齿轮箱故障工况下的容错控制运行,最后进行仿真实验,并将仿真结果与一个实际运行的2 MW风力发电机系统进行对比。实验结果表明:该故障检测与容错控制方案可以有效地对齿轮箱进行故障检测,并在存在故障时适当对风力发电机的输出功率进行降额控制,从而降低叶片和塔架承受的应力,并有效降低了齿轮箱轴承和润滑油的温度。

基于自注意机制胶囊网络的行星齿轮箱故障诊断

针对实际工程中行星齿轮箱故障数据有限、诊断准确率不高的问题,提出一种基于自注意机制胶囊网络的故障诊断方法。直接将采集到的行星齿轮箱振动信号作为输入,用首层宽卷积层提取浅层特征,过滤输入中的高频噪声;引入自注意机制关注信号关键特征;再次将所提特征输入胶囊层,进一步提取特征并实现故障分类;采用行星齿轮箱实验平台数据对所提方法进行实验验证。实验结果表明:在样本数量有限的情况下,所提方法仍能取得不错的诊断准确率。

变反激励力作用下齿轮箱减振降噪方法研究

从主动振动控制原理出发,通过改变反激励力的方式研究齿轮箱减振降噪效果。建立齿轮箱系统动力学模型,分析获取轴承支反力,采用有限元方法对齿轮箱进行振动特性分析,并建立箱体振动实验平台,结合有限元分析与实验测试结果确定次级力源布放位置;基于主动振动控制方法,在次级力源上施加不同激励力抑制其振动,分析在不同激励力作用下的减振效果;将获取的箱体表面振动数据作为边界条件导入LMS Virtual.Lab声学软件中计算齿轮箱向外声场辐射的噪声。研究结果表明:施加相位相反、振幅相同的反激励力减振效果最佳,相比于无控制状态平均位移减小0.18mm,施加抑制主导模态的反激励力时降噪效果最佳,在1191Hz处最高降噪量达到50.13dB。

WT2000系列风电机组齿轮箱空中更换技术研究

为了降低双馈异步风力发电机组主齿轮箱(增速箱)整体更换的成本,使作业流程更加简洁,在WT2000系列风电机组的基础上,设计一种主轴固定工装来临时代替齿轮箱的作用,固定主轴的后端。通过实践探究了WT2000系列风电机组齿轮箱空中更换作业中新的操作工艺和方法。结果表明:所设计工装大大降低了WT2000系列风电机组齿轮箱更换的作业成本(包含吊车成本和时间、人工成本),降低了维修期间的发电量损失。

复合多尺度注意熵在旋转机械多工况损伤识别中的应用

针对传统旋转机械损伤识别方法存在模型精度低和抗噪性差的问题,提出了一种基于复合多尺度注意熵(CMATE)和随机森林(RF)的旋转机械多工况损伤识别方法。首先,提出了一种新的测量时间序列复杂度的非线性动力学工具——复合多尺度注意熵;然后,利用CMATE提取旋转机械振动信号的损伤特征,其表征了旋转机械不同工况下的损伤特性;接着,将损伤特征输入至基于随机森林构造的多类别分类器中,进行了损伤识别;最后,采用滚动轴承-齿轮箱、齿轮箱和离心泵3种旋转机械数据集,并分别构造了9种工况和20种工况的多工况损伤数据集,对该损伤识别方法进行了实验研究。研究结果表明:该旋转机械损伤识别方法分别取得95%、97%和100%的识别准确率,在准确率和特征提取效率两方面优于其他的非线性动力学工具;并且在0 dB、1 dB、2 dB和3 dB这4种不同信噪比的噪声干扰下,依然取得了不错的损伤识别结果,证明了该模型具有可观的抗噪性;同时,该损伤识别方法能够稳定地识别旋转机械的不同负载和转速下的损伤,平均识别准确率分别达到了97.2%和93.5%,具有较好的实际应用潜力。

高速线材精轧锥辊箱故障诊断分析

针对某钢厂高速线材产线精轧区域锥辊箱振动问题,通过在线监测系统及时定位锥辊箱缺陷位置,准确判断劣化程度。对锥辊箱振动全过程进行了分析总结,并对在以后的振动分析时遇到类似的振动信号应考虑的问题提出一定见解。