基于匹配追踪时频LBP谱图的内燃机气门故障诊断方法

为直接对内燃机振动时频图像进行诊断识别,引入图像纹理特征提取技术,提出一种基于振动信号匹配追踪时频局部二值模式谱图的内燃机气门故障诊断新方法。首先,为清晰刻画内燃机时频图像中的各分量信息,利用匹配追踪算法(MP)获取无交叉项干扰项且时频聚集性良好的信号时频表示;然后引入局部二值模式(LBP)生成MP时频表示的LBP谱图,并将LBP谱图的灰度直方序列作为特征参数,利用支持向量机(SVM)对故障状态进行模式识别。在内燃机4种不同气门状态的诊断识别实验中,该方法最高识别率可达99.17%,表明基于MP时频LBP谱图识别的故障诊断方法具有较强的故障特征描述能力,能够准确诊断内燃机气门故障。

Method for IC engine valve fault diagnosis based on matching pursuit time-frequency LBP spectrogram

In order to directly diagnose and recognize vibration time-frequency images of IC engine, the image texture feature extraction technology was introduced and a novel method for IC engine valve fault diagnosis based on matching pursuit time-frequency local binary pattern (LBP) spectrogram was proposed. First, to clearly describe component information in time-frequency images of IC engine, the time-frequency representation of the signal with no cross-term and good time-frequency clustering was obtained by means of the matching tracking (MP) algorithm. Then LBP was introduced to yield LBP spectrogram of MP time-frequency representation. The gray histogram sequence of LBP spectrogram was taken as characteristic parameter, and support vector machine (SVM) was used for pattern recognition of fault state. In the diagnosis of four different states of IC engine valve, the maximum recognition rate of the proposed method reached 99.17%, which indicates that the proposed method based on MP time-frequency LBP spectrum identification has strong fault characterization ability and can accurately diagnose IC engine valve malfunction.