摘要
针对内燃机缸盖振动信号信噪比低且呈宽带、非平稳、时变特性,本文首先采用了基于软阈值的小波紧缩降噪方法进行处理,从而有效地突出了缸盖在不同时刻所受到激励的响应信号;然后,对降噪后的时域响应信号分别建立二阶AR模型,以模型参数作为模式识别特征参数;最后,采用感知器神经网络对特征参数进行识别与分类,由于缸盖振动信号信噪比的有效提高,使得对缸盖振动信号的时域识别和分类取得了很好的结果.通过对实验数据的处理,进一步验证了方法的有效性.
Recognition of cylinder head vibration signal of internal combustion engine in time domain is very important for on-line engine control, fault diagnosis and operation monitoring. Considering the low signal-noise ratio, broad-band, non-stationary and time-varying characteristics of the cylinder head vibration signal, in this paper a soft thresholding based wavelet shrinkage denoising approach is presented to process the signal at first. This effectively gives prominence to the response signals stimulated at different time on the cylinder head. Then, a second-order AR model is set up for the denoised response signals in time domain and the model' s parameters are taken as the pattern recognition characteristic parameters. Lastly, a perceptron neural network is used to recognize and classify these characteristic parameters. Since the signal-noise ratio of cylinder head vibration signal is increased effectively, satisfied result is obtained in the recognition and classification of cylinder head vibration signals in time domain. By the processing on experimental data, the validity of the proposed approach is certified.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2002年第2期242-245,共4页
Pattern Recognition and Artificial Intelligence