基于核密度估计和随机滤波理论的齿轮箱剩余寿命预测方法

针对风电机组齿轮箱的剩余寿命预测过程中需要进行状态退化模型结构假设的问题,提出一种由核密度估计和随机滤波理论结合的实时剩余寿命预测方法。该方法利用从数据本身出发的核密度估计方法对齿轮箱连续退化状态的概率密度函数进行非参数估计,得到齿轮箱实时状态监测数据的退化状态概率密度函数;利用实时状态监测数据来更新随机滤波递推模型参数,从而预测齿轮箱的实时剩余寿命。通过齿轮箱的试验验证了该方法的有效性。     

基于KDE的构架应力谱外推与疲劳寿命评估

获取动应力谱是对车辆结构进行疲劳分析的重要前提,其分析结果的准确性受动应力谱的完整性和充分性的影响。在保留载荷的概率分布特征的前提下,有必要将动应力测试数据从小样本外推至大样本。对转向架构架的应力谱外推和疲劳寿命评估进行了研究,并以某地铁转向架测点为分析案例。首先,对二维核密度估计（kernel density estimation,简称KDE）中最优带宽的确定方法进行了讨论;其次,提出了采用矩阵型灰色关联度分析方法对外推前后的雨流矩阵的接近性关联度与相似性关联度进行量化评价;最后,提出基于多样本核密度动应力谱外推的疲劳寿命评估方法。结果表明：与基于实测数据的评估结果相比,基于核密度外推的疲劳寿命评估结果的相对误差为-4.17%,安全运营里程减少了2.13×104 km。基于核密度估计的外推精度满足工程实践的要求,且评估结果更偏于安全,对保证车辆安全运营十分有益。     

基于核密度估计的实时剩余寿命预测

现有的许多设备由于自身故障样本数据不足、少有同类故障样本数据等,寿命预测研究时往往需要进行模型结构假设及参数估计。针对这类研究方法估计不够准确的问题,提出一种基于核密度估计的非参数实时剩余寿命预测方法。该方法利用能表征部件连续退化的特征量构建退化分布的核密度估计模型,进而得到剩余寿命的概率分布函数。在实时监测不断获得新的退化特征数据后,利用已知样本的核密度估计不断递推更新得到新增样本后的核密度估计,从而进一步实现对预测剩余寿命分布的更新。通过实例分析,验证了该方法在剩余寿命预测中的有效性。     

核密度估计法在板件概率损伤识别中的应用

基于谱元法将核密度估计方法用于解决结构的损伤识别,从而得到了损伤位置的概率密度函数。通过建立压电-结构耦合的三维谱元法模型,模拟Lamb波在铝板完好及损伤结构中的传播过程。利用连续小波变换计算响应信号在传感器之间的飞行时间,得到Lamb波在结构中的传播速度。通过对Rayleigh-Lamb方程的数值分析,得到Lamb波的理论传播速度,并将其与谱元法得到的结果进行对比,证实了谱元法模型的准确性。在椭圆定位技术的基础上,考虑环境不确定性对测量信号的影响,引入核密度估计方法将损伤位置识别转化为一种概率性问题。讨论了3种噪声水平情况下的损伤位置的概率密度函数,并给出了最终识别的结果。结果表明,核密度估计方法能够有效地识别出损伤位置,最大误差在5%左右。     

基于非参数核密度估计法的商用车服役载荷环境研究与应用

用户使用道路载荷特征的研究与建立是对车辆结构件疲劳强度合理轻量化设计的基础和前提。以往研究方法大多基于参数法概率密度估计对用户使用道路载荷特征进行统计分析。以某型商用车用户使用载荷特征的大范围数据调研为基础,分别采用参数法和非参数核密度估计方法,建立了典型用户使用道路载荷特征的概率密度分布模型,研究表明非参数核密度估计方法较之传统参数法更能准确表征用户使用载荷的分布特征。采用非参数核密度估计方法计算了90%分位下的各项用户使用道路载荷特征,结合实测用户道路载荷谱,基于非参数雨流矩阵外推方法构建了该型商用车减震弹簧应变的全寿命周期载荷,并依此载荷制定了台架疲劳试验规范。研究方法为其他结构件的用户关联疲劳试验规范的合理制定提供了一定的参考。     

基于局部振动测试的约束子结构模型修正方法

针对大型复杂结构模型修正中待识别参数过多、修正效率低等问题,提出一种基于局部振动测试的约束子结构模型修正方法。首先,利用子结构部分的振动测试响应构造约束子结构,从而将子结构从整体结构中隔离出来,形成独立简单的结构,有效地隔绝整体结构对子结构的影响。然后,基于构造的约束子结构响应数据,直接对待修正的目标子结构进行模型修正。最后,以一个T形钢管焊接结构和某发射台骨架为例,对所提方法的有效性进行了验证分析。结果表明,该方法不仅能够减少待识别参数,简化修正过程,而且修正效率及精度也较高,为大型复杂结构的模型修正提供了一种新思路。     

基于角结构统计量的MKECA间歇过程故障监测

针对间歇过程复杂非线性的特点,提出一种基于角结构统计量的多向核熵成分分析(MKECA)间歇过程监测方法。该方法首先将间歇过程数据进行标准化预处理,然后采用KECA提取间歇过程数据的主成分矩阵。研究表明,经过KECA投影后的主成分数据具有良好的角结构,因此利用主成分矩阵构造基于角结构的统计量,并且采用核密度估计算法计算其控制限。与传统的统计量相比,无需假设过程变量服从高斯分布。最后通过青霉素发酵的仿真平台和大肠杆菌实际生产过程验证,实验结果表明,相比于传统MKPCA方法,能够有效利用主成分的结构信息,明显降低了故障的误报率、漏报率。     

基于数据驱动方法的性能退化数据可靠性评估建模

采用退化量分布法对性能退化数据建立可靠性评估模型需要多个严格的假设前提,在实际应用中难以满足.为了克服退化量分布法的不足,提出一种数据驱动的退化数据可靠性评估模型.首先构建非参数自适应核密度估计得到各时刻产品退化量的概率密度函数,进而计算产品各时刻可靠度,最后通过三参数威布尔分布对各时刻可靠度进行拟合获得产品性能的可靠度曲线.基于实例数据,通过K-S检验证明了该模型相对参数方法具有稳定性和优良性,且不易冒进,便于工程实际应用.     

基于果蝇优化算法-小波支持向量数据描述的滚动轴承性能退化评估

针对支持向量数据描述(SVDD)算法对滚动轴承早期故障不敏感、参数选择困难的问题,提出了一种基于果蝇优化算法-小波支持向量数据描述(FOA-WSVDD)的滚动轴承性能退化评估方法。提取滚动轴承早期无故障振动信号的时域、时频域特征向量,并基于单调性进行特征选择;针对现有核函数对滚动轴承早期故障不敏感问题,将小波核函数引入到SVDD算法中;针对SVDD算法参数选择困难的问题,以支持向量个数与总样本数的比值作为适应度函数,采用改进的FOA算法对其核参数进行优化,建立FOA-WSVDD评估模型;最后,将轴承后期振动数据的特征向量输入到该WSVDD模型中,得到轴承的性能退化指标。试验结果表明,采用所提方法能准确地对轴承早期故障作出预警,与基于高斯核函数的SVDD算法相比,提前了17h。     

基于子结构分析的动态载荷和模型参数复合反演研究

针对结构特征参数、输入激励和输出响应等信息不完备的系统,基于子结构方法提出一种时域内动态载荷和模型参数复合反演的方法。首先,将结构划分成不同的子结构,对存在动态载荷的子结构建立有限元模型,该子结构与其他部件之间的连接关系采用等效的界面力代替。采用Green核函数法建立识别该子结构所受动态载荷的正问题模型,利用正则化方法实现时域内动态载荷的稳定识别。然后,在已识别载荷的基础上建立整体结构的有限元模型,通过非线性最小二乘法实现结构的未知模型参数的识别。数值算例表明,在响应存在测量噪声的情况下,所述的复合反演方法能够有效稳定地实现动态载荷和模型参数的复合反演。     

指数加权动态核主元分析法及其在故障诊断中应用

核主元分析法能充分利用核函数来解决非线性问题,具有很好的非线性逼近能力,但传统的核主元分析不能处理动态问题。在分析核主元分析法的基础上,提出一种新的指数加权核主元分析算法,建立一个多变量加权自回归统计核主元模型,选择Q统计量来判断系统是否发生故障,给出指数加权核主元分析法诊断故障的具体计算步骤。对液压泵进行了试验,利用小波包对液压泵端盖的振动信号进行处理,提取由13个时域和时频域特征量构成的故障特征矢量。试验结果表明,与传统的核主元分析法相比,新方法能实时更新主元模型和控制限Qa,合理地利用实时动态信息,能较好地处理动态问题,通过计算比较选择合适的加权因子,能获得良好的故障诊断效果,该方法是可行而有效的。     

基于估计点的双窗宽核密度估计算法

针对核密度估计中窗宽确定困难的问题,提出了基于估计点的滑动双窗宽核密度估计算法.通过采用固定窗宽的密度估计函数代替假设的正态分布密度函数以及增加求解二次导数的窗宽的方法,对估计域中的每一估计点求取其最优窗宽值,实现了窗宽根据样本的分布情况,在不同的估计点自动调整窗宽的取值.文中给出了算法的具体推导以及实现步骤,给出了多...     

A correlation consistency based multivariate alarm thresholds optimization approach

Different alarm thresholds could generate different alarm data, resulting in different correlations. A new multivariate alarm thresholds optimization methodology based on the correlation consistency between process data and alarm data is proposed in this paper. The interpretative structural modeling is adopted to select the key variables. For the key variables, the correlation coefficients of process data are calculated by the Pearson correlation analysis, while the correlation coefficients of alarm data are calculated by kernel density estimation. To ensure the correlation consistency, the objective function is established as the sum of the absolute differences between these two types of correlations. The optimal thresholds are obtained using particle swarm optimization algorithm. Case study of Tennessee Eastman process is given to demonstrate the effectiveness of proposed method.     

Process monitoring using kernel density estimation and Bayesian networking with an industrial case study

Principal component analysis has been widely used in the process industries for the purpose of monitoring abnormal behaviour. The process of reducing dimension is obtained through PCA, while T-tests are used to test for abnormality. Some of the main contributions to the success of PCA is its ability to not only detect problems, but to also give some indication as to where these problems are located. However, PCA and the T-test make use of Gaussian assumptions which may not be suitable in process fault detection. A previous modification of this method is the use of independent component analysis (ICA) for dimension reduction combined with kernel density estimation for detecting abnormality; like PCA, this method points out location of the problems based on linear data-driven methods, but without the Gaussian assumptions. Both ICA and PCA, however, suffer from challenges in interpreting results, which can make it difficult to quickly act once a fault has been detected online. This paper proposes the use of Bayesian networks for dimension reduction which allows the use of process knowledge enabling more intelligent dimension reduction and easier interpretation of results. The dimension reduction technique is combined with multivariate kernel density estimation, making this technique effective for non-linear relationships with non-Gaussian variables. The performance of PCA, ICA and Bayesian networks are compared on data from an industrial scale plant.     

面向响应准确度的参数不确定性模型确认方法

针对参数不确定性模型中概率与区间混合不确定性情况的模型确认问题,提出了面向响应准确度的模型确认方法,并阐明了确认过程的具体实施步骤。根据响应数值结果的非精确累积分布函数和响应试验结果的经验分布函数求解模型确认准则参数,并评估模型确认结果。对于不满足评估要求的模型,建立以待修正区间不确定性参数总置信水平最大为优化目标,以模型确认评估标准和模型参数初始值为约束条件的优化模型,运用遗传算法求解优化问题,并得到模型参数修正值。将模型确认方法应用在圣地亚热传导问题中,结果表明：所提方法可明显提高预测精度和响应准确度,结果真实可信。     

Improved reliability analysis method based on the failure assessment diagram

With the uncertainties related to operating conditions,in-service non-destructive testing(NDT) measurements and material properties considered in the structural integrity assessment,probabilistic analysis based on the failure assessment diagram(FAD) approach has recently become an important concern.However,the point density revealing the probabilistic distribution characteristics of the assessment points is usually ignored.To obtain more detailed and direct knowledge from the reliability analysis,an improved probabilistic fracture mechanics(PFM) assessment method is proposed.By integrating 2D kernel density estimation(KDE) technology into the traditional probabilistic assessment,the probabilistic density of the randomly distributed assessment points is visualized in the assessment diagram.Moreover,a modified interval sensitivity analysis is implemented and compared with probabilistic sensitivity analysis.The improved reliability analysis method is applied to the assessment of a high pressure pipe containing an axial internal semi-elliptical surface crack.The results indicate that these two methods can give consistent sensitivities of input parameters,but the interval sensitivity analysis is computationally more efficient.Meanwhile,the point density distribution and its contour are plotted in the FAD,thereby better revealing the characteristics of PFM assessment.This study provides a powerful tool for the reliability analysis of critical structures.     

改进 S 变换谱估计的动态手势肌电特征处理

手语是各种手势动态变化的一种复杂运动模式,手势特征处理效果直接关系到手语识别的准确性。本文提出一种基于改进S变换谱估计的动态手势肌电特征处理新方法。对采集的表面肌电信号进行S变换,引入优化因子调节时频分辨率并生成改进S变换谱;定义谱的时间和频率分量为二维随机变量,以改进S变换谱元素为二维随机变量样本,通过高斯核密度估计得到二维核密度函数。仿真和实验均表明,改进S变换谱估计方法有效抑制了白噪声,并使动态手势的肌电暂态突变特征得到加强。与经验模态分解、自排序熵、奇异值排序熵等方法对比,基于该方法的动态手势识别率分别提高了10.0%、6.67%和11.67%,特征处理方法的效果明显。     

Stereological analysis and modelling of gradient structures

Gradient structures are inhomogeneous along a particular gradient direction but homogeneous perpendicular to that direction. Consequently, structural parameters such as volume fraction or surface area density are local characteristics which depend on the 'vertical' coordinate with respect to the 'vertical' gradient axis.
Analogously, models for gradient structures have model parameters depending on the vertical coordinate z . For example, a Voronoi tessellation with a gradient is generated by a gradient point process with a local intensity which is a function of z . Similarly, a gradient germ grain model is obtained from a gradient point process where the grain size distribution may also depend on z . For a gradient Boolean model, local volume fraction V_V ( z ) and local surface area density S_V ( z ) can be calculated from the model parameters.
Stereological methods for gradient structures are based on vertical sections parallel to the gradient direction. Estimation of V_V ( z ), S_V ( z ) and local length density L_V ( z ) is done by lineal analysis using horizontal test lines with vertical coordinate z . Similarly, lineal analysis is used to estimate local mean cell volume of gradient tessellations. For the estimation of local particle number density and size in the spirit of the Wicksell problem the use of kernel methods and distributional assumptions is required.     

基于L0正则化模糊核估计的遥感图像复原

基于模糊图像的退化过程、卷积模糊模型和模糊图像生成的机理,提出一种基于L₀范数的正则化模糊核估计方法,解决了遥感图像重建问题中0范数难求解的难题。该方法以模糊核稀疏性为先验知识,采用对应梯度的L₀范数为正则项,有效避免了细小边缘对模糊核估计的影响,使得模糊核的估计更加准确。进一步采用超拉普拉斯分布来近似图像梯度的重尾分布,利用L_0.5范数正则化对模糊图像做反卷积,从而恢复出原始图像。与传统方法相比,本文方法可以准确地估计出图像的模糊核,很好地抑制恢复图像的振铃现象,有效地提升遥感图像的质量。模糊图像以及各方法重构图像在同一刀刃下的调制传递函数（MTF）曲线显示,本文方法的MTF曲线得到了较好的提升。     

Wavelet spectrum analysis approach to model validation of dynamic systems

Feature-based validation techniques for dynamic system models could be unreliable for nonlinear, stochastic, and transient dynamic behavior, where the time series is usually non-stationary. This paper presents a wavelet spectral analysis approach to validate a computational model for a dynamic system. Continuous wavelet transform is performed on the time series data for both model prediction and experimental observation using a Morlet wavelet function. The wavelet cross-spectrum is calculated for the two sets of data to construct a time-frequency phase difference map. The Box-plot, an exploratory data analysis technique, is applied to interpret the phase difference for validation purposes. In addition, wavelet time-frequency coherence is calculated using the locally and globally smoothed wavelet power spectra of the two data sets. Significance tests are performed to quantitatively verify whether the wavelet time-varying coherence is significant at a specific time and frequency point, considering uncertainties in both predicted and observed time series data. The proposed wavelet spectrum analysis approach is illustrated with a dynamics validation challenge problem developed at the Sandia National Laboratories. A comparison study is conducted to demonstrate the advantages of the proposed methodologies over classical frequency-independent cross-correlation analysis and time-independent cross-coherence analysis for the validation of dynamic systems.