列车轴承轨边声学诊断中故障声谱识别的时变阵列分析技术研究

发布时间：2018-06-18 01:44

本文选题：故障诊断 + 轨边声学　；参考：《中国科学技术大学》2017年博士论文

【摘要】：机械故障诊断技术的兴起,给高速发展中的各新兴制造业的安全保障提供了支持,减少和避免了重大事故的发生,表现出了巨大的经济和社会价值。高铁作为我国新兴装备制造业的代名词,其安全、舒适、高效一直受到国内外的广泛关注。因此,监测和诊断列车运行状态,预防列车事故的发生具有十分重大的意义。列车轮对轴承故障作为一种常见故障,与其相关的监测与诊断研究一直为国内外相关部门的热点。其中,轨边声学诊断系统以非接触式测量,可监测早期故障的特点而广受关注。然而,由于轨边声学诊断系统获取信号的独特方式,使其不可避免的存在一些测量问题。本文将以获取轨边声学诊断系统中清晰可辨的故障声谱为目标,以时变阵列分析为主要手段,针对系统测量所产生的声谱微弱、声谱畸变与声谱混叠三个问题进行探讨和研究,以期获取高速情况下可靠、准确的列车轴承诊断结果。论文首先通过分析轨边声学诊断系统的几何模型,公式化的揭示了该系统中声谱微弱、畸变、混叠三个测量问题产生的原因。针对目前主要使用的列车轴承,建立了以轴承故障频率为主要指标的故障类型判别基础。并以此分别设计了列车轴承静态声学获取方案和麦克风阵列的列车轴承动态声学获取方案。通过对静态和动态实验信号的频谱特征分析,验证了轨边系统中畸变混叠问题的存在。针对单麦克风分离矫正方法的局限性,以形态时频滤波与时频幅值匹配方法为例进行了深层次的分析探讨,指出了单麦克风信号对空间声源的欠定性描述是导致方法失效的本质原因。其次,针对声谱混叠问题,论文先后从远场条件下的阵列模型出发,提出了一条基于麦克风阵列的时变空域滤波重排的多源畸变混叠信号分离、矫正方案。该方法通过零角度空域滤波器获取不同声源的时间中心,并通过时变空域滤波重排最终实现不同畸变声源的分离与矫正。由于时变空域滤波器的建立与信号能量几乎无关,因此该方法在微弱信号源分离与矫正方面相比于传统单麦克风方法具有明显的优势。此外,实验表明所提方案对频带近似、时频能量离散的多源信号分离表现也十分良好。随后,针对声谱畸变问题,论文又对基于阵列的声谱畸变矫正方案进行了进一步的研究,并以单声源为例提出了一种基于时变多信号分类和角插值重采样的无参矫正方法。该方案通过时变多信号分类获取声源的实时位置,并通过声源发射时间和接收时间的一一映射关系建立重采样时间序列,实现对畸变信号的无参矫正。该方法相比于传统方法具有无需先验知识,计算量小,噪声鲁棒性强,适用于变速声源问题等诸多优点,在实际系统的应用中具有较高的潜力。最后,针对声谱特征微弱问题,论文以时变阵列分析思想为指导,通过汉克矩阵构建与阵列信号极其类似的伪阵列信号,提出了一种基于时变奇异值分解的周期性暂态信号声谱特征增强方法。重点研究了时变奇异值分解方法在处理周期暂态信号中所表现的基本性质,并以此及建立了一条基于时变奇异值分解的轴承故障频谱特征增强与识别路线。研究表明,该路线不仅在各类噪声,提高频谱特征信噪比表现优异,还在保留周期故障特征的谐波成分上具有较为显著的优势。通过对列车轴承的故障声学信号进行分析,表明该方案在提升故障信号的声谱特征方面作用明显。全文以麦克风阵列所采集的单、多声源轨边声学轴承故障信号为处理对象,从轨边声学信号采集模型与轨边信号的声谱特征出发,建立了一条完整的以时变阵列分析思路为主体的声谱混叠分离、声谱畸变矫正和声谱特征增强的技术路线,为最终实现列车轴承轨边声学系统故障声谱清晰识别提供了一定的研究基础和解决思路。
[Abstract]:The rise of mechanical fault diagnosis technology has provided support to the safety guarantee of the emerging manufacturing industry in high speed development, reducing and avoiding the occurrence of major accidents, showing great economic and social value. As the pronoun of the new equipment manufacturing industry of our country, the high speed railway has been widely concerned at home and abroad. Therefore, it is of great significance to monitor and diagnose the running state of the train and prevent the occurrence of train accidents. The train wheelset bearing fault is a common fault, and its related monitoring and diagnosis research has always been a hot spot in the relevant departments at home and abroad. However, because of the unique way of obtaining signals from the acoustic diagnosis system on the rail side, it inevitably has some measurement problems. This paper aims at obtaining the clear and distinguishable sound spectrum in the track edge acoustic diagnosis system, with the time-varying array analysis as the main hand, and the sound spectrum produced by the system measurement is weak and sound. Three problems of spectral distortion and sound spectrum mixing are discussed and studied in order to obtain reliable and accurate diagnosis results of train bearing in high speed conditions. Firstly, the paper analyzes the geometric model of the track acoustic diagnosis system, and formulae the reasons for the weak, distortion and aliasing of three measurement problems in the system. In order to use the train bearing, a fault type criterion based on the bearing fault frequency is established, and the dynamic acoustic acquisition scheme of the static acoustics of the train bearing and the dynamic acoustic acquisition scheme of the train bearing of the microphone array are designed respectively. The analysis of the spectrum characteristics of the static and dynamic experimental signals is used to verify the track edge system. In view of the limitation of the single microphone separation and correction method, the deep analysis is carried out with the morphological time frequency filtering and the time frequency amplitude matching method. It is pointed out that the underqualitative description of the single microphone signal to the spatial sound source is the essential reason for the failure of the method. Secondly, the theory of the sound spectrum aliasing is discussed. In this paper, based on the array model under the far field condition, a multi source distorted aliasing signal separation and correction scheme based on the time-varying spatial domain filter rearrangement based on the microphone array is proposed. This method obtains the time center of different sound sources through the zero angle spatial domain filter, and finally realizes the different distortion sound sources through the time-varying spatial domain filter rearrangement. Separation and correction. Since the establishment of the time-varying space filter is almost independent of the signal energy, this method has obvious advantages compared with the traditional single microphone method in the separation and correction of weak signal source. In addition, the experiment shows that the proposed scheme is very good for the frequency band approximation and the time frequency energy discrete multisource signal separation. Then, in order to solve the problem of acoustic spectrum distortion, this paper further studies the array based correction scheme of acoustic spectrum distortion, and presents a non parametric correction method based on time-varying multi signal classification and angular interpolation resampling with single sound source as an example. The scheme obtains the real-time position of the sound source through the time-varying multi signal classification and sends through the sound source. The resampling time series is established by the one-to-one mapping relationship between the shooting time and the receiving time to realize the non parametric correction of the distorted signal. Compared with the traditional method, the method has many advantages, such as without prior knowledge, small calculation, strong noise robustness, suitable for the problem of variable speed sound source and so on, and has high potential in the application of the actual system. Finally, the needle Based on the time-varying array analysis idea, the paper constructs a pseudo array signal which is extremely similar to the array signal by the Hank matrix, and proposes a method for the characteristic enhancement of the periodic transient signal based on the time-varying singular value decomposition. The time variant singular value decomposition method is focused on the processing of the transient transient signal. Based on the basic properties of the signal, a frequency spectrum feature enhancement and recognition route based on time-varying singular value decomposition is established. The study shows that the route is not only excellent in all kinds of noise, but also in improving the spectrum characteristic signal to noise ratio. After analyzing the fault acoustic signal of the train bearing, it shows that the scheme plays an obvious role in improving the sound spectrum characteristics of the fault signal. The full text is based on the single, multi source track edge acoustic bearing fault signal collected by the microphone array, and sets out from the sound spectrum characteristics of the acoustic signal acquisition mode and the edge signal of the rail. The technical route of the sound spectrum aliasing separation, the correction of the sound spectrum distortion and the enhancement of the sound spectrum features by the time-varying array analysis idea provides a certain research basis and solution for the final realization of the clear recognition of the fault sonogram of the train bearing rail edge acoustic system.
【学位授予单位】：中国科学技术大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：U279.323

【相似文献】