基于振动信号的车用发动机运行状态预测

CONDITION PREDICTION OF VEHICLE ENGINE BASED ON VIBRATION SIGNALS

考虑到车用发动机结构复杂、振源较多,振动信号为非平稳信号的特点,文中利用混沌与分形理论对多组不同运行状态的振动序列进行研究,探讨关联维数与运行状态之间的内在联系,并利用混沌与神经网络相结合的方法对主要状态参量进行单变量及多变量预测。实验结果表明,关联维数能敏感反应发动机的磨损状态,而多变量的预测效果比单变量效果理想。

As the vehicle engine is a complex structure with various vibration resources and the picked-up vibration signals are usually non-stationary, chaos and fractals theory is used to study a series of vibration signals, which are acquired under different running states in order to discover the intrinsic relationships between the correlation dimension and the engine states. Based on the chaos and neural network theory the prediction of the main state parameters is realized through the single variable time series and the multi-variable ones respectively. The experiment results show that the correlation dimension is sensitive to the wear conditions and the multi-variable prediction is better than the single variable one.