【作者】 张晓晓；

【导师】 刘志刚；

【作者基本信息】 西南交通大学 ， 轨道交通电气化与自动化， 2013， 硕士

【摘要】 受电弓和接触网之间的接触压力波动已成为衡量接触网不平顺程度及弓网受流性能的重要指标。本文以接触压力信号为研究对象,采用现代信号处理中的各种谱分析方法对其进行深入地统计分析,从而尝试建立一种合适的接触网线谱来评价接触线的不平顺状态及衡量弓网系统的性能。本文以实测京广线及沪昆线接触压力数据(时速100km/h-200km/h)为研究对象,对其平稳性问题、时域分析、频域分析及时频域分析方面进行了研究,主要内容如下：(1)随机信号的平稳性判定是对其进行统计处理的重要前提。首先对信号平稳性检验领域中存在的各种方法进行了比较全面的综述；然后针对接触压力信号的平稳性问题,分别利用轮次检验,单位根检验,基于替代数据法、基于谱分析的平稳性检验方法对京广线与沪昆线的实测接触压力数据(采样频率为2个／米)进行了分析,综合各检验结果对其平稳性做了相应的判定：接触压力信号的平稳性是相对于观测尺度而言的,在较长的采样距离下(采样里程约1000m及以上),被测线路的动态接触压力波动可基本视为平稳随机过程；对于较短的采样距离下(采样里程500m以下),一些接触压力数据段出现明显的非平稳性。因此在对接触压力信号进行统计分析时,应选取合适的处理方法。另外本文提出接触压力信号非平稳度的概念,作为评价线路的参数设计、弓网匹配及接触线平顺状态等的参考指标。(2)利用时域分析技术求取了接触压力信号的各项时域指标,根据各指标的变化趋势,从时域角度分析了接触压力数据的统计特性,并借以简单评价弓网的受流状态以及接触线平顺状态。(3)利用小波分析理论和经验模式分解理论,结合接触压力数据本身特点,提出了识别其波长成分的方案。首先对接触压力信号进行经验模式分解,去除固有模态分量中的伪分量、低频分量及残差项,得到包含100m以下波长成分的接触压力信号；然后利用小波变换对该信号进行多层分解得到逼近信号及细节信号,实现信号中的长短波成分分离；最后对得到的逼近信号与细节信号分别进行频谱及时频谱分析。实验分析表明：本方案可以有效地识别接触压力信号中的固有波长以及瞬态突变波长成分；接触压力信号的细节信号时频谱可以作为评估弓网受流性能,识别及定位接触线病害的参考。(4)利用高阶统计量中的谱峭度方法对接触压力信号进行了统计分析,并与小波变换进行结合。实验分析表明：基于小波变换的谱峭度对接触压力信号中的非高斯非平稳成分有着很强的表征能力,可为评价接触线不平顺及弓网受流性能提供一种参考；另外提出利用小波变换谱峭度方法对接触压力信号进行平稳性检验的思路。更多还原

【Abstract】 The fluctuation of the contact force has already been regarded as an index that evaluates the contact line unevenness degree and the current collection performance between the pantograph and the catenary. In recent years, constructing catenary spectrum has become a new idea to evaluate the overall dynamic performance of catenary and pantograph system. In the paper, the contact force signals are analyzed with the modern signal processing methods.For the stationarity of contact force and the statistics characteristic in time domain and frequency domain, the work is listed as follows:1) Testing the stationarity of the random signal is an important preprocessing step before studying its statistical feature. Firstly, the methods on testing the stationarity of the signals are fully summarized. Then based on the data of Beijing-Guangzhou line and Shanghai-Kunming line(two data points sampled per mile), the stationarity of contact force is analyzed by both using the methods in frequency domain, which compares the time-frequency spectrum similarity of the signal, such as surrogates-based test and spectrum-based test, and using the methods in time domain, such as the run test and a series of unit root tests. The testing results show that, whether the contact force signal can be considered as stationary, relatively to a given observation scale. From a longer observation scale (sample distance is over1000m), the contact force sections can be consider as stationary; from a small observation scale (sample distance is below500m), the dynamic contact force appears obvious non-stationary. At last, a non-stationarity index is proposed as a quantitative tool to assess the fluctuations of the contact force.2) The contact force is analyzed by the time domain indexes which are applied to initially evaluate the unevenness degree of the contact line and monitor the state of pantograph and contact lines.3) Based on wavelet transform theory and empirical mode decomposition, the contact force wavelength identification procedure is presented. As a result, the wavelength composition in normal sections and abnormal sections are identified and located. Firstly, the contact force is pretreated by using EMD and the signal concluding the wavelength ranging from lm to100m is obtained. Secondly, the pretreated contact force is decomposed into multilayer detail signals and approximation signals via the wavelet transform, then the long and short wavelength ingredients are separated. Lastly, the detail signals and the approximation signals are respectively analyzed in the frequency domain and time-frequency domain. The result implies that the combination of EMD and the wavelet transform could identify the natural wavelength and mutation wavelength effectively and accurately; the contact force detail signal time-frequency spectrum may be as a reference to evaluate the performance of the current collection and the contact line fault identified.4) Based on the data of Beijing-Guangzhou line and Shanghai-Kunming line, the spectral kurtosis based on wavelet transform are applied to the contact force signal processing. The results show that, the spectrum kurtosis based on wavelet transform has the ability of detecting the local mutation in the contact force, which may be as a reference to evaluate the current collection performance; the idea that testing the stationarity of the contact force via the spectral kurtosis based wavelet transform is proposed.更多还原