论文部分内容阅读
基于柴油机气缸盖振动信号,利用rbf神经网络建立振动信号与压力之间的映射,从而通过训练好的网络实现压力重构。为了剔除噪声干扰,在时域采用同步平均法对信号进行降噪;在频域,通过相干分析确定各频率成分相干性,据此设计滤波器进行滤波。在经过时域、频域两次降噪后信噪比得到较大提高,提高了重构精度。通过在6135型柴油机上的实验证明,该方法简单有效,重构精度高,具有较强的可操作性。
Based on the diesel engine cylinder head vibration signal, the mapping between vibration signal and pressure is established by rbf neural network, so that the pressure reconstruction can be realized through the trained network. In order to eliminate the noise interference, the signal is denoised in the time domain by using the synchronous averaging method. In the frequency domain, the coherence of each frequency component is determined by coherent analysis, and the filter is designed accordingly. After two time-frequency and frequency-domain noise reduction, the signal-noise ratio is greatly improved and the reconstruction accuracy is improved. The experiment on 6135 diesel engine proved that the method is simple and effective, high reconstruction accuracy and strong operability.