柴油机运行时因激振力的作用会产生一定方向和频率的冲击振动,构件的裂纹或松动等故障会影响到其响应成分的频率能量特性。针对柴油机运行时的冲击响应振动信号,利用小波分析快速进行信噪分离,频域范围内采用功率谱分析结合小波包分解对各频段能量谱分析。根据振动信号时域峰值和时刻,频域能量的变化和分布,给出故障诊断层使用的状态特征向量,并用比例梯度动量共轭算法训练的神经网络模型进行柴油机状态定位与故障识别。
Diesel engine operation will generate vibration with certain direction and frequency due to the role of exciting force.Cracks of parts and components or loose parts and components will affect the frequency and energy of their responses to the action.Wavelet analysis can be used in the live-fire vibration signals to get the signal noise separation quickly.Power spectrum analysis with wavelet packet can decompose the energy spectrum of the frequency band.A state feature vector used for fault diagnosis is given according to the peak and moment of time domain vibration signal,as well as change and distribution of frequency domain energy.Momentum conjugate gradient algorithm for the proportion of trained neural network model is put forward for diesel engine positioning and fault identification.