高速列车非平稳振动信号盲源分离方法及应用

高速列车具有若干时变激励源,传统的时频分析方法只能对观测的混合振动的总体强度分布、时频域结构加以分析,不能分离出与各振源对应的信号分量从而明晰振源状态与故障特征。盲源分离是一种可行的分析方法,但由于高速列车振动信号具有时变振源数目、时变信号长度、受车速调制的变频非平稳等特征,传统的盲源分离方法不适用。为了提高高速列车非平稳信号的盲源分离效果,基于自适应滤波理论提出全局最优信噪比盲源分离新方法,并对其可分离性的判别依据进行论证。新方法的有效性经仿真计算和实测数据分析得到验证。研究表明：新方法对高速列车时变非平稳信号的盲源分离效果优于传统的基于非线性函数的盲源分离方法和基于高阶累积量的盲源分离方法。     

混合信号振声源分离与多机组故障诊断

针对机房设备混合信号难以提取有用信息,提出了多参数的振声诊断方法。应用最小互信息梯度下降的盲分离算法,通过展开边缘熵和修正四阶累积量估计值的方法改善算法性能,在故障源数量未知且可能大于传感器数量的情况下,根据信息源之间的独立性测度关系依次提取最显著的特征值。仿真结果证明,改进算法估计误差减小且算法可靠。在诊断实例中,首先,分离机房内的混合噪声信号以确定主要故障来源;然后,采集故障源的振动信号进行非线性盲分离,提取热泵机组压缩机不对中、齿轮啮合不良和碰磨的故障特征;最后,根据分离的振源信号特征识别故障类型,建立基于盲源分离算法的大空间设备群的振声诊断方法。     

强脉冲干扰非线性混叠转子故障信号分离

由于旋转机械在运行过程中,传感器测得的振动信号是各振源的混叠信号且含有很强的噪声,常规的信号处理方法难以分离混叠信号,对设备的状态监测和故障诊断造成了很大的困难。针对这一问题,介绍了盲源分离基本原理和方法,指出源分离算法在脉冲噪声环境下失效。针对强脉冲噪声环境下的混叠振动信号,首先,通过中值滤波降噪方法对振动信号进行降噪;然后,通过盲源分离算法对降噪后的信号分离;最后,利用该方法对实测混叠转子振动信号成功实现了降噪和故障信号分离。仿真结果验证了提出方法的有效性。     

Application of the blind source separation method to feature extraction of machine sound signals

As the result of vibration emission in air, a machine sound signal carries important information about the working condition of machinery. But in practice, the sound signal is typically received with a very low signal-to-noise ratio. To obtain features of the original sound signal, uncorrelated sound signals must be removed and the wavelet coefficients related to fault condition must be retrieved. In this paper, the blind source separation technique is used to recover the wavelet coefficients of a monitored source from complex observed signals. Since in the proposed blind source separation (BSS) algorithms it is generally assumed that the number of sources is known, the Gerschgorin disk estimator method is introduced to determine the number of sound sources before applying the BSS method. This method can estimate the number of sound sources under non-Gaussian and non-white noise conditions. Then, the partial singular value analysis method is used to select these significant observations for BSS analysis. This method ensures that signals are separated with the smallest distortion. Afterwards, the time-frequency separation algorithm, converted to a suitable BSS algorithm for the separation of a non-stationary signal, is introduced. The transfer channel between observations and sources and the wavelet coefficients of the source signals can be blindly identified via this algorithm. The reconstructed wavelet coefficients can be used for diagnosis. Finally, the separation results obtained from the observed signals recorded in a semianechoic chamber demonstrate the effectiveness of the presented methods.     

Application of the blind source separation method to feature extraction of machine sound signals

As the result of vibration emission in air, a machine sound signal carries important information about the working condition of machinery. But in practice, the sound signal is typically received with a very low signal-to-noise ratio. To obtain features of the original sound signal, uncorrelated sound signals must be removed and the wavelet coefficients related to fault condition must be retrieved. In this paper, the blind source separation technique is used to recover the wavelet coefficients of a monitored source from complex observed signals. Since in the proposed blind source separation (BSS) algorithms it is generally assumed that the number of sources is known, the Gerschgorin disk estimator method is introduced to determine the number of sound sources before applying the BSS method. This method can estimate the number of sound sources under non-Gaussian and non-white noise conditions. Then, the partial singular value analysis method is used to select these significant observations for BSS analysis. This method ensures that signals are separated with the smallest distortion. Afterwards, the time-frequency separation algorithm, converted to a suitable BSS algorithm for the separation of a non-stationary signal, is introduced. The transfer channel between observations and sources and the wavelet coefficients of the source signals can be blindly identified via this algorithm. The reconstructed wavelet coefficients can be used for diagnosis. Finally, the separation results obtained from the observed signals recorded in a semi-anechoic chamber demonstrate the effectiveness of the presented methods .     

A subspace method for the blind extraction of a cyclostationary source: Application to rolling element bearing diagnostics

The need for blindly separating mixtures of signals arises in many signal processing applications. A class of solutions to this problem was recently proposed by the so-called blind source separation (BSS) techniques which rely on the sole knowledge of the number of independent sources present in the mixture. This paper deals with the case where the number of sources is unknown and statistical independence may not apply, but where there is only one signal of interest (SOI) to be separated, which is cyclostationary. It proposes a blind extraction method using a subspace decomposition of the observations via their cyclic statistics. This method is first developed for instantaneous mixtures and is then extended to the convolutive case in the frequency-domain where it does not suffer from the permutation problem as does classical BSS. Experiments on industrial data are finally performed and illustrate the high performance of the proposed method.     

A repeated single-channel mechanical signal blind separation method based on morphological filtering and singular value decomposition

This paper proposes a repeated blind source separation (BSS) method based on morphological filtering and singular value decomposition (SVD) to separate the mixed sources from a single-channel signal. Firstly the signal is de-noised by the morphological filter and, the noise which affects the accuracy of the separation is removed. Next, the purified signal is reconstructed in phase space, and the SVD is applied to this matrix. After choosing the effective singular values, the inverse transform is applied to the revised signal matrix. From this, the pseudo signal can be obtained. The pseudo signal and the purified original signal are used to achieve the mixed sources separation through the fast independent component analysis (FastICA) algorithm. Then, the methods above are repeated in order to separate the weaker signals. The analysis of simulation and practical application demonstrates that that proposed method shows a high level of separating performance of a single-channel signal.     

周期平稳信号盲源分离算法及其应用

周期平稳信号盲源分离方法是在没有先验知识的前提下,从一组采集信号中提取未知源周期平稳信号。对混合的周期平稳信号进行盲源分离的算法进行研究。根据循环平稳度的特征,提出以循环平稳度作为盲源分离准则,取其旋转过程的极大值点,即为循环平稳度最大时,分离出周期平稳信号源。在理论推导的基础上,对分离准则进行证明。通过仿真验证,把该方法应用于齿轮箱振动数据实例分析中,表明循环平稳度作为分离准则有很好的分离未知源周期平稳信号的效果。     

基于阶比滤波的单通道缸盖振动信号盲源分离

传统盲源分离方法要求传感器观测信号数目不小于源信号数目,且在源信号平稳、相互独立的前提下,才能得到较为准确的分离信号,但对于发动机缸盖振动非平稳信号,由于激励源较多,这些条件不易满足。为实现缸盖振动信号盲源分离,提出了基于阶比滤波的单通道缸盖振动信号盲源分离方法。利用燃爆激励信号频率随转频变化的先验信息,通过阶比滤波得到阶比分量,将阶比分量和单通道信号组成多维观测信号,通过快速独立成分分析方法得到了缸盖振动非平稳信号的分离信号。仿真和应用研究证明了该方法的有效性。     

基于多特征融合的行星齿轮箱点蚀故障诊断研究

为了识别行星齿轮箱的齿面点蚀故障,通过刚柔耦合仿真获得健康和 3 种不同点蚀程度行星齿轮箱的箱体振动信号。对获得的 4 种状态的箱体振动信号进行变分模态分解后,计算每个本征模态函数分量的能量值、峭度因子和信息熵,基于能量值、峭度因子和信息熵多特征融合构建高维特征向量,采用支持向量机分类器对 4 种状态的行星齿轮箱进行识别。结果表明,基于变分模态分解的本征模态函数分量的能量值、峭度因子和信息熵构建的 15 维特征向量,采用支持向量机分类器能够准确识别健康和 3 种不同点蚀程度齿轮的类型。     

滚动轴承故障检测的改进包络分析法

提出了一种基于连续小波变换和谱峭度分析的改进包络分析方法,并将其应用于滚动轴承的故障检测之中。该方法的核心是自动确定由轴承缺陷所引起的共振频率所在的频带,自动构建最佳包络来进行故障诊断,该方法已成功地应用到了对故障轴承振动信号的仿真和实测,并取得了满意的结果。     

Blind equalization using combined skewness–kurtosis criterion for gearbox vibration enhancement

In this paper a method for vibration signal enhancement is presented. It incorporates an idea that the signal acquired on the machine housing is a convolution of an informative signal (cyclic pulse train) with an impulse response of the system. The impulse response corresponds to a transmission path through which the informative signal propagates. The informative signal is a signal that contains information about a local damage. The classical method that estimates the impulse response of the system is called minimum entropy deconvolution (MED) and it aims to maximize kurtosis of the deconvolved signal, i.e. kurtosis of the informative signal estimate. Recently, skewness-based deconvolution (equalization) has been proposed as an alternative method for damage detection in rotating machines. In this paper we incorporate an alternative criterion which combines advantages of both of the previously used deconvolution criteria. Kurtosis is a widely-used tool for impulsiveness detection even if they are hidden in the signal, although favouring single-spike signals is a disadvantage of kurtosis. On the other hand, skewness is more robust, since it incorporates statistical moment one order lower than kurtosis. However, signals related to local damage are not always asymmetric, thus skewness is not a suitable criterion for their extraction. Thus, it is worth to combine both kurtosis and skewness in a single deconvolution criterion. We compare properties of two previously used criteria (kurtosis and skewness) with the novel one which is based on the Jarque–Bera statistic using a simulation study. An experimental validation on a real vibration signal (two-stage gearbox from an open-pit mine) is performed as well.     

Separation of fault features from a single-channel mechanical signal mixture using wavelet decomposition

This study reports a joint wavelet decomposition and Fourier transform approach to the separation of periodic mechanical source signals from single-channel signal mixture. With this method, the signal mixture is first decomposed to certain wavelet scales. The resulting wavelet coefficients are then Fourier transformed to extract the information pertinent to each signal source from these scales. Next, the number of signal sources is determined and the wavelet coefficients for each signal are constructed in all scale levels. Finally the source signals can be reconstructed using these wavelet coefficients. Since this method does not require the number of sources to be known a priori, it is particularly suitable for mechanical fault signal separation as the number of source signals varies with time and is unpredictable. It is also important to point out that the number of sources is determined without the commonly adopted sequential extraction/learning process and hence the proposed method can be used for on-line fault detection due to the reduced computing burden. The application of this method has been demonstrated using mixed bearing data containing both inner and outer race fault signals.     

Simulating gear and bearing interactions in the presence of faults: Part I. The combined gear bearing dynamic model and the simulation of localised bearing faults

This paper presents a simulation model for a gearbox test rig, in which a range of bearing faults can be implemented. Bearing faults sometimes manifest themselves by their interaction with meshing gears, and to simulate this it is necessary to model a whole system of gears and shafts supported in bearings. This has now been done for an experimental test rig through the incorporation of a time-varying, non-linear stiffness bearing model into a previously developed gear model. The incorporated bearing model is based on Hertzian contact theory, which relates the raceway displacement to the bearing load, and also accounts for the slippage between the elements. It has the capacity to model localised spalls (inner race, outer race and rolling elements), which are discussed in this part of the paper and extended inner and outer race faults (rough surfaces), which are discussed in part II. Even though the whole gearbox has not been modelled in detail, the non-linear time-varying gear-meshing operation is modelled in some detail. Both simulated and experimental localised fault signals (acceleration signals) were subjected to the same diagnostic techniques; namely spectrum comparisons, Spectral Kurtosis (SK) analysis and envelope analysis. The processed simulated signals showed a similar pattern to that observed in their measured counterparts and were found to have a characteristic, referred to in the literature as double pulses, corresponding to entry into and exit from the localised fault. The simulation model will be useful for producing typical fault signals from gearboxes to test new diagnostic algorithms, and possibly prognostic algorithms.     

Principal component analysis and blind source separation of modulated sources for electro-mechanical systems diagnostic

Under the only hypothesis of independent sources, blind source separation (BSS) consists of recovering these sources from several observed mixtures of them. As it extracts the contributions of the sources independently of the propagation medium, this approach is usually used when it is too difficult to modelise the transfer from the sources to the sensors. In that way, BSS is a promising tool for non-destructive machine condition monitoring by vibration analysis. Principal component analysis (PCA) is applied as a first step in the separation procedure to filter out the noise and whiten the observations. The crucial point in PCA and BSS methods remains that the observations are generally assumed to be noise-free or corrupted with spatially white noises. However, vibration signals issued from electro-mechanical systems as rotating machine vibration may be severely corrupted with spatially correlated noises and therefore the signal subspace will not be correctly estimated with PCA.This paper extends a robust-to-noise technique earlier developed for the separation of rotating machine signals. It exploited spectral matrices of delayed observations to eliminate the noise influence. In this paper, we focus on the modulated sources and prove that the proposed PCA is available to denoise such sources as well as sinusoidal ones. Finally, performance of the algorithm is investigated with experimental vibration data issued from a complex electro-mechanical system.     

LMD-ICA联合降噪方法在滚动轴承故障诊断中的应用

针对经典独立分量分析（ICA）只能应用于观测源数不少于信号源数的超定盲源分离问题,提出局部均值分解和ICA相结合的欠定盲源分离新方法。该方法将采集的单通道振动信号进行局部均值分解,基于互相关准则对分解的分量进行重组,构建虚拟噪声通道;将虚拟噪声通道与振动信号作为盲源分离的信号输入,采用基于负熵的FastICA算法实现信号源和噪声的分离,从而达到降噪目的。将该方法应用于滚动轴承故障信号,频谱分析结果表明,该方法处理后的信号中噪声得到一定程度滤除,频谱中毛刺更少,故障特征频率更加明显,有利于故障特征的提取,实验分析证明了该方法的有效性。     

Local fault detection in helical gears via vibration and acoustic signals using EMD based statistical parameter analysis

Gear is a vital transmission element, finding numerous applications in small, medium and large machinery. Excessive loads, speeds and improper operating conditions may cause defects on their bearing surfaces, thereby triggering abnormal vibrations in whole machine structures. This paper describes the implementation of empirical mode decomposition (EMD) method for monitoring simulated faults using vibration and acoustic signals in a two stage helical gearbox. By using EMD method, a complicated signal can be decomposed into a number of intrinsic mode functions (IMF) based on the local characteristic time scale of the signal. Vibration and acoustic signals are decomposed to extract higher order statistical parameters. Results demonstrate the effectiveness of EMD based statistical parameters to diagnose severity of local faults on helical gear tooth. Kurtosis values from EMD and that obtained from vibration and acoustic signals are compared to demonstrate the superiority of EMD based technique.     

Processing Human Colonic Pressure Signals by Using Overdetermined ICA

Independent component analysis (ICA) is a widely used method for blind source separation (BSS).The mature ICA model has a restriction that the number of the sources must equal to that of the sensors used to collect data,which is hard to meet in most practical cases.In this paper,an overdetermined ICA method is proposed and successfully used in the analysis of human colonic pressure signals.Using principal component analysis (PCA),the method estimates the number of the sources firstly and reduces the dimensions of the observed signals to the same with that of the sources;and then,FastICA is used to estimate all the sources.From 26 groups of colonic pressure recordings,several colonic motor patterns are extracted,which not only prove the effectiveness of this method,but also greatly facilitate further medical researches.     

Noise cancellation of ultrasonic NDE signals using blind source separation

This paper proposes a new denoising method for ultrasonic NDE (nondestructive evaluation) signals using blind separation (BSS) technology. The proposed denoising method consists of four steps. First, a reconstructed phase space (RPS) is constructed from observed ultrasonic NDE signals. The information about the underlying sources (e.g., ultrasonic signal, noise, etc.) acting on this system is contained in this RPS. Second, independent component analysis (ICA) is performed on the RPS to recover all sources underlying the RPS. Next, the ultrasonic signal component is selected by a decision criterion related to the denoising application and, finally, is reconstructed to obtain the denoised ultrasonic signal. To validate the proposed method, it has been applied to the experimental ultrasonic NDE signals of the test sample and is compared with the wavelet denoising method in SNR (signal-to-noise ratio) enhancement. The experimental results show that the SNR of the ultrasonic NDE signals can be enhanced greatly using the proposed denoising method and the proposed method has almost the same denoising performance as the wavelet denoising method in SNR enhancement. A trait of the proposed denoising method is the ability to denoise ultrasonic NDE signals by separating the ultrasonic signal and noise using blind source separation technology. The text was submitted by the authors in English.     

一种适用于信源数时变的自适应盲源分离算法

针对信源数动态变化情况下的盲源分离问题,首先采用一种基于交叉验证技术的估计方法用于估计时变的源数,然后推导了一种基于自然梯度和Frobenius范数相结合的自适应盲源分离算法,该算法不需要对源信号作任何约束性的假设,因此该算法适合于分离服从超高斯或亚高斯分布的信号。提出的算法通过了源数不变和源数动态变化2种方式实验的验证。