多抽样率分析在柴油机气缸压力神经网络识别中的研究

Multirate Analysis on Neural Network Recognition of Diesel Cylinder Pressure

利用柴油发动机缸盖振动信号识别气缸压力,对于柴油机故障诊断及运行状态监测是一个重要的环节,近年来,此类研究引起了相关学科的极大兴趣.在利用径向基神经网络对柴油机气缸压力的识别中牵涉到在一个周期内的采样点数的不同,极大地影响了在不同转速下识别气缸压力的适应性.因此,提出多抽样率分析,即根据缸盖振动信号的频带宽度,通过对原始信号的插值和抽取改变抽样频率,实现不同抽样点数的神经网络运算.

It is important in fault diagnosis and condition monitoring of diesel engine to recognize the cylinder pressure wave of diesel engine based on the vibration signal of cylinder head. The study has recently brought about great interests in related subjects. The number of samples in a working cycle under different rotating speeds, even in the same velocity, is different. This greatly affects the recognition of the cylinder pressure using a radial based function neural network. A multirate analysis method is proposed to solve the problem, i.e., resampling data by interpolating and decimating the original signal based on the bandwidth of the cylinder head vibration signal to realize the neural network arithmetic under different number of sampling points.