Adaptive autoregressive modeling of non-stationary vibration signals under distinct gear states. Part 2: experimental analysis

Parametric time-frequency representation based on parametric models is more desirable for presenting highly precise time-frequency domain information due to its high-resolution property. However, the sensitivity and robustness of parametric models, in particular the parametric models on the basis of advanced adaptive filtering algorithms, has never been investigated for on-line condition monitoring of rotating machinery. Part 1 of this study proposed three adaptive parametric models based on three advanced adaptive filtering algorithms. Part 2 of this study is concerned with the effectiveness of the proposed models under distinct gear states, especially the highly non-stationary conditions accrued from advanced gear faults. Four gear states are considered: healthy state, adjacent gear tooth failure, non-adjacent gear tooth failure and distributed gear tooth failure. The vibration signals used in this study include the time-domain synchronous averaging signal and gear motion residual signal for each considered gear state. The test results demonstrate that the optimum filter behavior can readily be attained and the white Gaussian assumption of innovations can relatively be easily guaranteed for the NAKF-based model under distinct gear states and a wide variety of model initializations. On the other hand, the EKF- and MEKF-based models are capable of generating more accurate time-frequency representations than the NAKF-based model, but in general the optimality condition for white Gaussian assumption cannot be guaranteed for these two advanced models. Therefore, the NAKF-based model is preferred for automatic condition monitoring due to its appealing robustness to distinct gear states and arbitrary model initializations, whereas the EKF- and MEKF-based models are desirable when accurate time-frequency representation is concerned.     

New techniques of local damage detection in machinery based on stochastic modelling using adaptive Schur filter

Vibration signal analysis is one of the most effective techniques of monitoring machinery and detecting local damage in their parts, e.g. bearings and gearboxes. However, such detection is sometimes difficult, especially in heavy industrial machines, because of a small proportion of damage-induced components in relation to the remaining components of registered signals. Therefore, more effective signal processing algorithms are being looked for. Moreover, local damage (cracking, pitting, spalling, breakage, etc.) in bearings and gearboxes generates broad-spectrum impulse signals, while the other type can be effectively modelled as a sum of narrowband signals. In this article, techniques based on Schur adaptive filter are proposed for local damage detection. In such an approach, the analysed signal is modelled by means of autoregressive process and the filter is described by so-called reflection coefficients. Schur algorithm is an effective algorithm with very good numerical properties and it is capable of tracking rapid changes in second order statistics of the analysed signal. Thus, the method is well-suited to analysing non-stationary signals and it is potentially interesting for use in bearing and gearbox monitoring.Reflection coefficients describing the signal model, defined with the use of Schur algorithm, may be applied in a variety of ways, giving a chance of employing different solutions in different conditions. In the first proposed solution, detection is based on the weighted sum of derivatives of reflection coefficients, while in the other one – on new signal obtained as power in frequency bands calculated from a parametric spectrogram, whose starting point are reflection coefficients. All these operations are aimed at enhancing changes that occur in the signal at the moments when damage-induced impulses appear. The article also presents guidelines for methods of determining parameter values in the employed analyses. The proposed solutions have been applied for analysing signals coming from a two-stage gearbox of a large machine driving a mining belt conveyor and the obtained results were analysed. They prove the effectiveness of the proposed techniques. It is worth emphasizing that these techniques can be easily adapted for monitoring machinery in varying operating conditions.     

Estimation of the blood Doppler frequency shift by a time-varying parametric approach

Doppler ultrasound is widely used in medical applications to extract the blood Doppler flow velocity in the arteries via spectral analysis. The spectral analysis of non-stationary signals and particularly Doppler signals requires adequate tools that should present both good time and frequency resolutions. It is well-known that the most commonly used time-windowed Fourier transform, which provides a time-frequency representation, is limited by the intrinsic trade-off between time and frequency resolutions. Parametric methods have then been introduced as an alternative to overcome this resolution problem. However, the performance of those methods deteriorates when high non-stationarities are present in the Doppler signal. For the purpose of accurately estimating the Doppler frequency shift, even when the temporal flow velocity is rapid (high non-stationarity), we propose to combine the use of the time-varying autoregressive (AR) method and the (dominant) pole frequency. This proposed method performs well in the context where non-stationarities are very high. A comparative evaluation has been made between classical (FFT based) and AR (both block and recursive) algorithms. Among recursive algorithms we test an adaptive recursive method as well as a time-varying recursive method. Finally, the superiority of the time-varying parametric approach in terms of frequency tracking and delay in the frequency estimate is illustrated for both simulated and in vivo Doppler signals.     

A Kalman filter algorithm based on exact modeling for FOG GPS/SINS integration

The Kalman filter is widely applied in fiber optic gyro (FOG) inertial integrated navigation system. To solve the problem of hard acquirement of Kalman filter parameters, a novel algorithm for FOG GPS/SINS integration navigation based on exact modeling is proposed in this paper. The models of inertial sensors using Allan variance analysis are established in proposed algorithm and the precise Kalman filter model is obtained based on the correspondence between Allan variance coefficients and inertial sensors parameters. The simulation and experimental results show that Kalman filter parameters can be obtained for GPS/SINS integrated navigation system precisely and efficiently based on Allan variance modeling method, and the algorithm has reference value for theoretical perfection and engineering applications.     

基于改进雁群PSO算法的模糊自适应扩展卡尔曼滤波的SLAM算法

针对扩展卡尔曼滤波同时定位与地图创建算法中难以建立准确的先验噪声模型的问题, 提出一种基于改进雁群粒子群算法的模糊自适应卡尔曼滤波算法. 利用分数阶微积分改进粒子进化速度, 利用混沌来改进粒子的初始化和发生早熟时的处理. 改进后的雁群粒子群算法在收敛速度与避免早熟方面有了很大改进, 并将改进的雁群粒子群算法用于模糊自适应扩展卡尔曼滤波同时定位与地图创建算法的训练, 并与用雁群粒子群算法训练的模糊自适应扩展卡尔曼滤波同时定位与地图创建算法进行对比, 其在定位与构图方面有很大的提高. 关键词： 同时定位与地图创建 雁群粒子群算法 分数阶微积分 混沌     

Flexible polarization demultiplexing method based on an adaptive process noise covariance Kalman filter

A flexible polarization demultiplexing method based on an adaptive Kalman filter(AKF) is proposed in which the process noise covariance has been estimated adaptively. The proposed method may significantly improve the adaptive capability of an extended Kalman filter(EKF) by adaptively estimating the unknown process noise covariance. Compared to the conventional EKF, the proposed method can avoid the tedious and time consuming parameter-by-parameter tuning operations. The effectiveness of this method is confirmed experimentally in 128 Gb/s 16 QAM polarization-division-multiplexing(PDM) coherent optical transmission systems. The results illustrate that our proposed AKF has a better tracking accuracy and a faster convergence(about 4 times quicker)compared to a conventional algorithm with optimal process noise covariance.     

Gearbox vibration monitoring using extended Kalman filters and hypothesis tests

The efficiency of many maintenance programs is heavily dependent on the detection accuracy of the condition monitoring system. Condition indicators that are sensitive to environmental or operational variables of no interest will inevitably reflect irrelevant fluctuations and thus mislead the subsequent analysis. In consideration of this phenomenon, a fully automatic and robust vibration monitoring system for gearboxes is proposed in this study. The primary objective here is on how to exclude the effects of variable load conditions. The proposed technique features a number of appealing advantages, which include extended Kalman filter-based time-varying autoregressive modeling, automatic autoregressive model order selection with the aid of a non-paired two-sample Satterthwaite's t′-test, a highly effective and robust condition indicator (the means of one-sample Kolmogorov-Smirnov goodness-of-fit test), and an automatic alert generating mechanism for incipient gear faults with the aid of a Wilcoxon rank-sum test. Two sets of entire lifetime gearbox vibration monitoring data with distinct variable load conditions were used for experimental validation. The proposed condition indicator was compared with other well-known and/or recently proposed condition indicators. The results demonstrate excellent performance of the proposed technique in four aspects: the effectiveness of identifying the optimum model order, a minimum number of false alerts, constant behavior under variable load conditions, and to some extent an early alert for incipient gear faults. Furthermore, the proposed condition indicator can be directly employed by condition-based maintenance programs as a condition covariate for operational maintenance decision analysis. It provides a quantitative and more efficient means for exchanging condition information with maintenance programs in comparison with the widely used non-parametric time-frequency techniques such as wavelets, which rely on visual inspection.     

Dispersion-based short-time Fourier transform applied to dispersive wave analysis

Although time-frequency analysis is effective for characterizing dispersive wave signals, the time-frequency tilings of most conventional analysis methods do not take into account dispersion phenomena. An adaptive time-frequency analysis method is introduced whose time-frequency tiling is determined with respect to the wave dispersion characteristics. In the dispersion-based time-frequency tiling, each time-frequency atom is adaptively rotated in the time-frequency plane, depending on the local wave dispersion. Although this idea can be useful in various problems, its application to the analysis of dispersive wave signals has not been made. In this work, the adaptive time-frequency method was applied to the analysis of dispersive elastic waves measured in waveguide experiments and a theoretical investigation on its time-frequency resolution was presented. The time-frequency resolution of the proposed transform was then compared with that of the standard short-time Fourier transform to show its effectiveness in dealing with dispersive wave signals. In addition, to facilitate the adaptive time-frequency analysis of experimentally measured signals whose dispersion relations are not known, an iterative scheme for determining the relationships was developed. The validity of the present approach in dealing with dispersive waves was verified experimentally.     

Modal parameter identification through Gabor expansion of response signals

A new method of identifying modal parameters by decomposing response signals with Gabor transform is presented in this paper to estimate natural frequencies, damping ratios and mode shapes of linear time invariant systems. According to Gabor expansion theory, responses of a multi-degree-of-freedom system can be decomposed into uncoupled signal components, each vibrating at a single natural frequency. From these uncoupled signals, modal parameters are subsequently extracted with common methods. The proposed method can process stationary and non-stationary responses and requires no input signal except for the response signals generated by unknown excitation acting on a system. In the sense of less restriction on the in-out signals, the approach based on time-frequency decomposition is very general. A simulation study on a simply supported beam under non-stationary excitation has demonstrated that the proposed method is effective in parameter estimation.     

基于卡尔曼滤波的低复杂度去混响算法*

在电话会议、智能音箱等应用场景下,传声器往往处在声源的远场。混响信号的存在会掩蔽后续到达的直达声信号,降低传声器接收信号的语音质量,以及语音识别系统的准确识别率。多通道线性预测算法是一种经典的盲去混响算法,但该算法往往具有较高的计算复杂度。本文提出了一种简化的卡尔曼滤波更新算法,通过对角化卡尔曼滤波器状态向量误差协方差矩阵,降低了自适应多通道线性预测去混响算法的复杂度。通过与现有分块对角简化算法对比发现,本文提出的简化算法在保证语音质量的同时,进一步降低了原卡尔曼滤波算法的复杂度。     

Speech enhancement based on the discrete Gabor transform and multi-notch adaptive digital filters

This paper presents a new method to speech enhancement based on time-frequency analysis and adaptive digital filtering. The proposed method for dual-channel speech enhancement was developed by tracking frequencies of corrupting signal by the discrete Gabor transform (DGT) and implementing multi-notch adaptive digital filter (MNADF) at those frequencies. Since no a priori knowledge of the noise source statistics is required this method differs from traditional speech enhancement methods. Specifically, the proposed method was applied to the case where speech quality and intelligibility deteriorate in the presence of background noise. Speech coders and automatic speech recognition (ASR) systems are designed to act on clean speech signals. Therefore, corrupted speech signals by the noise must be enhanced before their processing. The method uses a primary input containing the corrupted speech signal while a reference input containing the noise only. In this paper, we designed MNADF instead of single-notch adaptive digital filter and used DGT to track frequencies of corrupting signal because fast filtering process and fast measure of the time-dependent noise frequency are of great importance in speech enhancement process. Therefore, MNADF was implemented to take advantage of fast filtering process. Different types of noises from Noisex-92 database were used to degrade real speech signals. Objective measures, the study of the speech spectrograms and global signal-to-noise ratio (SNR), segmental SNR (segSNR), Itakura-Saito distance measure as well as subjective listing test demonstrated consistently superior enhancement performance of the proposed method over traditional speech enhancement method such as spectral subtraction. Combining MNADF and DGT, excellent speech enhancement was obtained.     

Sequential Bayesian geoacoustic inversion for mobile and compact source-receiver configuration

Geoacoustic characterization of wide areas through inversion requires easily deployable configurations including free-drifting platforms, underwater gliders and autonomous vehicles, typically performing repeated transmissions during their course. In this paper, the inverse problem is formulated as sequential Bayesian filtering to take advantage of repeated transmission measurements. Nonlinear Kalman filters implement a random-walk model for geometry and environment and an acoustic propagation code in the measurement model. Data from MREA/BP07 sea trials are tested consisting of multitone and frequency-modulated signals (bands: 0.25-0.8 and 0.8-1.6 kHz) received on a shallow vertical array of four hydrophones 5-m spaced drifting over 0.7-1.6 km range. Space- and time-coherent processing are applied to the respective signal types. Kalman filter outputs are compared to a sequence of global optimizations performed independently on each received signal. For both signal types, the sequential approach is more accurate but also more efficient. Due to frequency diversity, the processing of modulated signals produces a more stable tracking. Although an extended Kalman filter provides comparable estimates of the tracked parameters, the ensemble Kalman filter is necessary to properly assess uncertainty. In spite of mild range dependence and simplified bottom model, all tracked geoacoustic parameters are consistent with high-resolution seismic profiling, core logging P-wave velocity, and previous inversion results with fixed geometries.     

Robust Blind Adaptive Channel Equalization in Chaotic Communication Systems

Based on the bounded property and statistics of chaotic signal and the idea of set-membership identification, we propose a set-membership generalized least mean square （SM-GLMS） algorithm with variable step size for blind adaptive channel equalization in chaotic communication systems. The steady state performance of the proposed SM-GLMS algorithm is analysed, and comparison with an extended Kalman filter （EKF）-based adaptive algorithm and variable gain least mean square （VG-LMS） algorithm is performed for blind adaptive channel equalization. Simulations show that the proposed SM-GLMS algorithm can provide more significant steady state performance improvement than the EKF-based adaptive algorithm and VG-LMS algorithm.     

逼近积分点数下限的五阶容积卡尔曼滤波定轨算法

为了在保持滤波定轨精度不变的条件下提高定轨计算的实时性,提出一种新的逼近积分点个数下限的五阶容积卡尔曼滤波定轨算法.首先,采用一种数值容积准则对非线性函数的高斯加权积分进行近似,该准则所需的积分点个数仅比五阶代数精度容积准则积分点个数的理论下限多一个积分点,并在贝叶斯滤波算法框架下推导出本文算法的更新步骤.然后,给出实时定轨所需的状态方程和量测方程,在状态方程中考虑了J2项引力摄动和大气阻力摄动,在量测方程中利用坐标系转换推导了轨道状态与测量元素之间的非线性关系.仿真实验结果表明,本文所提算法在定轨精度方面与已有的五阶滤波算法相当,但所需的积分点个数最少,计算实时性最高,从而验证了本文算法的有效性.     

扩展Kalman滤波器同时测定苯酚和邻氯苯酚

提出了苯酚和邻氯苯酚混合体系非线性吸光度表示式。即非线性吸光度由两组分各自对吸光度的贡献、由于混合而导致的两组分各自贡献的改变以及交互作用项三部分构成。依据该式,利用非线性Kalman滤波紫外分光光度法同时测量了苯酚和邻氯苯酚混合体系。在250～290 nm 区间测量了32组浓度各自在1～15 mg·L-1范围的苯酚和邻氯苯酚标准混合溶液紫外吸收光谱图。利用偏最小二乘法,将其制作成扩展Kalman滤波标准工作系数矩阵。通过对非线性吸光度关系式Taylor级数展开进行线性化处理,得到其向量函数的Jacobi矩阵,从而完成了非线性Kalman滤波器的设置。回收实验显示出,扩展Kalman滤波同时测量苯酚和邻氯酚双组分体系是准确、稳定的。     

Time-resolved tracking of a sound scatterer in a complex flow: nonstationary signal analysis and applications

It is known that ultrasound techniques yield nonintrusive measurements of hydrodynamic flows. For example, the study of the echoes produced by a large number of particles insonified by pulsed wavetrains has led to a now-standard velocimetry device. In this paper, a new technique for the measurement of the velocity of individual solid particles moving in fluid flows is proposed. It relies on the ability to resolve in time the Doppler shift of the sound scattered by the continuously insonified particle. For this signal-processing problem two classes of approaches can be used: time-frequency analysis and parametric high-resolution methods. In the first class the spectrogram and reassigned spectrogram is considered, and applied to detect the motion of a small bead settling in a fluid at rest. In nonstationary flows, methods in the second class are more robust. An approximated maximum likelihood (AML) technique has been adapted, coupled with a generalized Kalman filter. This method allows for the estimation of rapidly varying frequencies; the parametric nature of the algorithm also provides an estimate of the variance of the identified frequency parameters.     

Bivariate empirical mode decomposition and its contribution to wind turbine condition monitoring

Accessing difficulties and harsh environments require more advanced condition monitoring techniques to ensure the high availability of offshore wind turbines. Empirical mode decomposition (EMD) has been shown to be a promising technique for meeting this need. However, EMD was developed for one-dimensional signals, unable to carry out an information fusion function which is of importance to reach a reliable condition monitoring conclusion. Therefore, bivariate empirical mode decomposition (BEMD) is investigated in this paper to assess whether it could be a better solution for wind turbine condition monitoring. The effectiveness of the proposed technique in detecting machine incipient fault is compared with EMD and a recently developed wavelet-based ‘energy tracking’ technique. Experiments have shown that the proposed BEMD-based technique is more convenient than EMD for processing shaft vibration signals, and more powerful than EMD and wavelet-based techniques in terms of processing the non-stationary and nonlinear wind turbine condition monitoring signals and detecting incipient mechanical and electrical faults.     

Tracking noisy limit cycle oscillation with nonlinear filters

For engineering systems, the dynamic state estimates provide valuable information for the detection and prediction of failure due to noise and vibration. From this perspective, nonlinear filtering techniques are applied to the problem of state estimation of dynamical systems undergoing noisy limit cycle oscillation. Specifically, the extended Kalman filter, ensemble Kalman filter and particle filter are used to track the limit cycle oscillations of a Duffing oscillator using noisy observational data. The noisy limit cycle oscillations feature highly non-Gaussian trends. The efficiency and limitations of the extended Kalman filter, ensemble Kalman filter and particle filter in tracking limit cycle oscillations are examined with respect to the model and measurement noise and sparsity of measurement data. For the limit cycle oscillations considered here, it is demonstrated that the ensemble Kalman filter and particle filter outperform the extended Kalman filter in the presence of sparse observational data or strong measurement noise. For moderate measurement noise and frequent measurement data, the ensemble Kalman filter and particle filter perform equally well in comparison to the extended Kalman filter.     

Multi-parameter identification from scalar time series generated by a Malkus-Lorenz water wheel

We address the issue of multi-parameter estimation from scalar outputs of chaotic systems, using the dynamics of a Malkus water wheel and simulations of the corresponding Lorenz-equations model as an example. We discuss and compare two estimators: one is based on a globally convergent adaptive observer and the second is an extended Kalman filter (EKF). Both estimators can identify all three unknown parameters of the model. We find that the estimated parameter values are in agreement with those obtained from direct measurements on the experimental system. In addition, we explore the question of how to distinguish the impact of noise from those of model imperfections by investigating a model generalization and the use of uncertainty estimates provided by the extended Kalman filter. Although we are able to exclude asymmetric inflow as a possible unmodeled effect, our results indicate that the Lorenz-equations do not perfectly describe the water wheel dynamics.     

光谱导数Kaiman滤波同时测定苯酚-邻氯苯酚和2，4二氯苯酚

文中提出了一个新的、稳定的光谱导数Kalman滤波紫外分光光度法同时定量分析方法，并且成功地将其应用于极难分辨的苯酚一邻氯苯酚和2，4二氯苯酚三组分混合体系的同时测量。选择分析光谱导数的原因是其不仅包含着吸光度，还包括在指定波长位置变化趋向等更多的信息量，这样更利于在吸收峰重叠的位置捕捉有较大差异的信号。利用Kalman滤波可以解决由光谱导数而引起的噪声放大问题，也可以过滤源于实验的噪声以及来自传递模型误差。在260-290nm区间测量了30组浓度各自在1～10mg·L^-1范围的苯酚一邻氯苯酚和2，4二氯苯酚标准混合溶液紫外吸收光谱图。利用分段延伸高次多项式回归模拟求导准确获得吸光度导数，并且利用偏最小二乘法将其制作成Kalman滤波标准工作系数矩阵。通过随机线性离散系统的Kalman滤波器最优平滑计算，对混合体系中的各组分进行定量分析。回收实验显示出，光谱导数Kalman滤波分析法应用于本实验的极难分辨的三组分体系，得到了非常高的回收率，并且在整个实验范围内光谱导数Kalman滤波分析法具有非常好的稳定性。