采用非线性动力学方法研究脑精神疾病是近年来国内外学者研究的热点和趋势.针对脑精神疾病的研究和诊断中缺少客观有效的量化参数和量化指标的状况,提出了一种根据对时间序列功率谱划分而定义的谱熵,然后用其计算和分析脑电信号谱熵的方法.通过数据仿真试验证明该谱熵和信号活跃性之间存在正相关关系.基于这种相关性,应用该方法对抑郁症患者和正常对照组的脑电信号功率谱熵进行了数值计算,然后进行了分析对比和统计检验.实验结果表明:抑郁症患者脑电信号的功率谱熵在部分脑区显著弱于正常健康人.证明该谱熵能够表征大脑电生理活动状况,提供反映其活动性强弱的信息,可以作为度量大脑电生理活动性的一个参数.这对于能否将该功率谱熵作为诊断脑精神疾病的物理参数具有积极意义.

A method is proposed to calculate and analyze electro-encephalogram signal to improve the situation that there is an urgent need for an effective quantitative indicator to describe brain mental disorders. The method defines a spectral entropy in terms of the power spectrum division of time series. Then, the entropy is applied to numerical calculation of electroencephalogram signals of depression patients and normal control group. Meanwhile, the differences are compared between them. Experimental results show that the power spectral entropy in depression patients is significantly weaker than the normal healthy people’s in some brain regions. Further analysis proves two facts. One is that the entropy is positively correlated to brain electrical physiological activity, and the other tells that the entropy can be used as a parameter to measure brain electrical activity, to characterize brain electrical physiological activities, and to provide the activity intensity information. This paper

针对轴承振动信号中存在周期性冲击这一现象,提出了时间-小波能量谱熵的计算方法,用于滚动轴承的故障诊断。首先构造脉冲小波,采用连续小波变换的方法得到时间域内小波能量谱,再沿时间轴计算能量谱熵,定量描述振动信号沿时间的分布情况,不同故障下轴承的冲击振动随时间变化程度不同,其时间-小波能量谱熵值也就不同。将不同故障轴承信号的时间-小波能量谱熵作为向量特征输入建立支持向量机,实现了对轴承的工作状态和故障类型的判断。实验结果表明,时间-小波能量谱熵可以有效地对滚动轴承进行故障诊断。

Aiming at periodic impulses in vibration signals of bearings,a new method so called time-wavelet energy spectrum entropy was proposed for rolling element bearing fault diagnosis.Firstly,an impulse response wavelet was constructed to obtain wavelet energy spectrum in time domain by using the continuous wavelet transformation,then the energy spectrum entropy describing quantitatively vibration signals varying with time was calculated along the time axis, the impact vibrations of bearings with different faults varied with time at different levels,and their time-wavelet enery spectrum entropies were different.To identify the fault pattern and the working condition of a bearing,the time-wavelet enery spectrum entropies of different fault signals could be taken as input vectors of a support vector machine.Practical examples showed that the proposed method can be used to diagnose efficiently faults of rolling element bearings.

针对实测水泵振动信号夹杂噪声等现象,利用多尺度数学形态谱及形态谱熵理论,对不同转速下的水泵振动实测信号进行分析,提取相应的特征形状和特征量,进行振动故障识别,并验证了该方法的可行性与准确性。研究表明,随着转速的逐步增加,存在着不对中和不平衡故障的水泵机组,其振动特性可以划分为正常运行、周期性碰摩运行及混沌运行3类;当水泵振动类型相同时,随着转速的变化,相应的数学形态谱及形态谱熵分析结果几乎不变;采用数学形态谱及形态谱熵理论能够对夹杂大量噪声的故障测试信号进行分析处理,且故障信号识别成功率较高,其中形态谱识别成功率高于形态谱熵,能较好地达到故障识别的目的,也充分证明了该方法的抗干扰性。

In view of observed pumps vibration signals with noises, the multi-scale mathematical morphology spectrum and spectrum entropy are used to analyze the signals with different revolving speeds. The features of signals are extracted to recognize the vibration types, and then the feasibility and accuracy of the method are proved. The results show that with the gradually-in-creased speed, vibration features of the pump units with misalignment and coupling unbalance fault can be divided into three types, namely normal operation, periodic rubbing and chaotic operation. When the vibration type is same, the analysis results based on morphological spectrum and spectrum entropy are almost the same for any revolving speeds change. The observed signals with plenty of noise can be analyzed and handled effectively by the two methods, however the recognition success rate of morpho-logical spectrum is higher, which contributes to vibration recognition.

基于相关函数的常规功率谱估计方法在脉冲噪声环境中性能下降明显,受大脉冲干扰时甚至会失效。而具有局部相似度量特性的相关熵对脉冲信号的影响并不敏感,能够很好地适应脉冲噪声条件。将受干扰信号的相关熵函数代替自相关函数,得到基于相关熵的功率谱估计方法。通过仿真进行对比验证,结果表明在相同信噪比条件下基于相关熵的功率谱估计方法在α稳定分布噪声环境中更加稳健,特别在大脉冲条件下(α1.6)能对有用信号进行更加有效的估计,且估计误差的均值和方差值仍保持很低。

Conventional power spectrum density (PSD)estimation method based on correlation function is obviously de-graded in the impulsive noise environment,even loses efficacy under big impulsive noise.As a new localized similarity measure,correntropy is not sensitive to the impulsive noise and adapts this environment.Correntropy spectrum density (CSD)estimation method is achieved through correntropy function replacing autocorrelation function of signal.With the same SNR the CSD method is robust to the PSD method through simulation.Specially,the CSD method is also effective un-der big impulsive noise (α<1.6),and the mean and variance of estimation error remain low level.

针对基于时域组合特征的故障诊断方法的不足，提出一种基于小波包能谱熵分析的液压油缸内泄漏故障诊断方法。分析无杆腔压力信号的时域特征，采用小波包变换提取压力信号的能谱熵并输入到改进LM神经网络进行内泄漏的故障诊断。实验结果表明，无泄漏压力信号的能谱熵向量各元素分布较均匀；而泄漏信号的能谱熵向量各元素差异较大；改进 LM 神经网络在精度、准确率等方面高于传统BP、LM 神经网络。与时域组合特征法进行比较，结果验证算法的高效可检测性。以不同分类器、不同小波基对算法诊断性能的影响进行分析，结果表明，该方法具有很强的稳定性和优越性。

In view of the disadvantage of the fault diagnosis method based on time domain feature in which the required lighter leakage is similar to no leakage in time domain .A new fault diagnosis method is proposed .This method analyzes the time domain feature which is extracted from the pressure signal of hydraulic cylinder ,and then ,a wavelet packet decomposition method is adopted to extract the energy spectrum entropy from pressure signal at different scales .Lastly ,the energy spectrum entropy is input into the optimized LM neural network analyzer to identify no leakage , slighter leakage and heavy leakage .Experimental results show that each element of the spectrum entropy vector of normal signal of no leakage is evenly distrituted ,while the element of fault signal spectrum entropy vector is remarkably regularly changed . The accuracy and precision of the optimized LM neural network are higher than those of the traditional neural network model .The proposed method is compared to method