【作者】 孔庆鹏；

【导师】 陈鹰； 宋开臣；

【作者基本信息】 浙江大学 ， 机械电子工程， 2006， 博士

【摘要】 由于车辆工业的发展和对乘用车舒适性的要求，提出了从内部结构方面进行设计优化来减小发动机振动和噪声的要求。发动机特征分析是对发动机进行优化设计、故障诊断和状态监测的重要组成部分，而振动测试方法又是特征分析的主要手段。如何通过对外部测量振动信号进行特征信息提取，从而建立起特征信息和机器内部结构的关系，是非常有意义的。本课题是研究从复杂的发动机振动信号中提取和转速相关的信号，分离干扰噪声的阶比跟踪方法。工程师们可以利用所提取的阶比信号得到发动机的特征信息，更好地理解发动机内部结构，从减小发动机的运转振动和噪声方面对发动机的优化设计进行指导。在阶比跟踪分析中，准确地获得基准转轴的转速是非常关键的。针对在实际应用中转速计脉冲信号测量不准的情况，直接计算脉冲到达时间得到的转速曲线误差较大。本论文提出了分段加权三次多项式曲线拟合方法，通过对转速曲线进行分段拟合，通过数据重叠和加权平均来保证各分段点处的连续及平滑，得到准确的转速曲线。当前应用最为广泛的等角度重采样阶比跟踪方法一般是通过过采样插值实现等角度重采样的，最大信噪比只能达到65dB。本论文根据采样信号复原原理，提出了用插值滤波器来实现自适应数字重采样的方法。用一个预先设定好的具有很高过采样率的内插sin c(t)函数构成的滤波器实现了等角度自适应重采样，提高了自适应重采样的精度和阶比谱的信噪比。采用等角度重采样加FFT进行阶比跟踪，只能得到三维瀑布图或二维阶比谱图和阶比跟踪谱，并不能得到阶比的时域提取。本论文研究了Gabor变换及满足信号时域重构的对偶函数双正交条件，实现了在时频域进行可变中心频率带通滤波来分离阶比和噪声，并采用相似对偶函数最小二乘法一次迭代实现了阶比成分的时域重构信号。成功地实现了阶比在时域的波形提取。发动机结构复杂，是由许多轴的旋转和往复运动组成。不同转轴对应的阶比在频率上会有交叉，使阶比跟踪变得困难。等角度重采样加FFT方法只能以一个参考轴作为重采样的参照，提取的阶比波形就会受到相交叉阶比的干扰，产生失真。本文在Vold-Kalman自适应滤波器的基础上进行了改进，设计一个通带中心频率随着转速变化而变化的带通滤波器。用最小二乘法结合Kalman自适应滤波器同时提取多轴阶比，解决了多轴交叉阶比提取的耦合问题。设计了几种不同幅值变化的仿真信号，对上述三种阶比跟踪方法进行了比较。设计了阶比跟踪实测实验方案，利用Visual C++和Matlab混合编程编制了实验测试软件，搭建了发动机测试平台和信号采集系统。通过气动发动机台架自测数据与美国提供实验数据相结合，分别采用不同的阶比跟踪算法对不同来源的发动机实测数据进行了阶比跟踪实验研究。最后，对本文的研究工作做了简单的总结，提出了研究的不足之处和进一步研究的建议。发动机变速阶段振动信号阶比跟踪研究更多还原

【Abstract】 With the development of automobile industry and demand of riding comfort of car, design optimization from interior structure to lower engine vibration and noise is needed. How to extract characteristic information from exterior measurement vibration signal and get interior structure information is of vital meaning.This project is to research the order tracking methods to extract rotation related signal and separate disturbed noise from complicated engine vibration signal. Engineers can better understand engine interior structure through extracted order signal, and instruct optimization design from depressing running vibration and noise aspect.In order tracking analysis, it is very important to acquire rotating speed of reference shaft. In practical application, tachometer impulse measurement is not accurate. Therefore rotating speed curve through direct computing impulse arriving time may have big error. Aiming at this problem, a subsection weighing cubic spline curve fit method was proposed in this paper. This proposed method uses cubic spine curve fit rotating speed subsection, and through data overlap and weighing average to ensure continuuty and smooth of subsection point. Accurate rotating speed was achieved.Wildly applied even-angle re-sampling order tracking method is commonly implemented by over sampling data and interpolation, and can only get 65dB signal noise ratio. Base on sampled signal recover theory, an interpolation filter adaptive numeric re-sampling method was proposed in this paper. Adaptive even-angle re-sampling is implemented by a filter which is constructed by pre-established over sampling rate interpolation sin c(t) function. Accuracy of re-sampling andsignal-noise-ratio of order spectrum were improved.Even-angle re-sampling plus FFT order tracking method can get waterfall, order spectrum and order tracking spectrum result view, but it can not get order extraction in time domain. Gabor transform and dual function bi-orthogonal condition to fulfill time domain reconstruction were researched in this paper. Vary center frequency band pass filter in time-frequent domain was implemented to separate order and noise. And order component reconstruction in time domain was achieved by similar duel function least square method iteration. Order component was extracted in time domain by Gabor transform.Engine is complicated in structure, and is consists of rotating and reciprocating movement of many shafts. Orders of different shafts crossing each other in frequency make order tracking difficult. Even-angle re-sampling plus FFT method can only take one shaft as re-sampling reference, extracted order waveform be interrupted by crossing order and cause distortion. A few changes was done on Vold-Kalman adaptive filter, and a band pass filter with center frequency vary with rotating speed was designed. Least square method combined with Kalman adaptive filter can extract orders of multi-shaft, solves crossing order decouple problem.Several simulation signals with different amplitude change were designed. Comparison was done between three order tracking methods. A field experiment scheme was also designed. Test software was programmed by combining Visual C++ and Matlab. Engine test platform and signal collection system were constructed. Expermental data were collected form air powered engine platform and provided by Dactron corperation. Field expremet was carried out by different order tracking methods.Final, a brief summarization of research job was done, and shortcoming and research proposal were put forward.更多还原