谱减在基于动载的路面不平度识别中的应用  

Application of spectral subtraction in road roughness recognition based on dynamic load

作　　者：李忠国[1] 张为公[1] 周耀群[1] 冯长钰[1] 

机构地区：[1]东南大学仪器科学与工程学院,南京210096

出　　处：《东南大学学报（自然科学版）》2007年第1期51-55,共5页Journal of Southeast University：Natural Science Edition

基　　金：江苏省交通厅资助项目(05c02)

摘　　要：分析了路面激励和发动机振动引起的汽车垂直动载的特点,提出了采用谱减法去除用于路面不平度识别的垂直动载中周期性噪声的方法.谱减之前对染噪信号和噪声模板都作静态小波滤波,可有效消除随机噪声谱引起的“音乐噪声”.对发动机怠速时和在光滑水泥路面测得的信号比较后采用了后者作为噪声模板,并使用基本谱减方法对信号进行了滤波.试验数据处理发现谱分辨率取2Hz时谱估计方差较小,相应“音乐噪声”也较少.仿真表明,在数据点数为1 024个时,使用谱减去噪使信噪比由7.2dB提高到31.5dB.Following performance analysis of vehicle vertical dynamic load （VVDL）resulted from road inspiritment and engine vibration, spectral subtraction method is proposed to eliminate the cyclic noise in VVDL used in road roughness recognition. Before spectral subtraction, stationary wavelet filtering is made for noisy signal and noise model, which can effectively eliminate ＂music noise＂ caused by random noise spectral. After comparing VVDL signal measured when engine idling with that obtained on smooth cement road, the latter is adopted as the noise model and the basic spectral subtraction method is employed to filter signals. When spectral resolution is 2 Hz, variance of spectral estimation is smaller and less ＂music noise＂ would be yielded after spectral subtraction. Simulation indicates signal to noise ratio （SNR） is improved from 7.2 dB to 31.5 dB with spectral subtraction method when there are 1 024 data samples.

关 键 词：滤波 谱减 汽车振动 车轮力 路谱 

分 类 号：U467.4[机械工程—车辆工程]