Elman人工神经网络对发动机缸壁间隙的识别诊断

Diagnosis Identification of Engine Cylinder Wall Clearance based on Elman Artificial Neural Network

主要以DA462型发动机作为研究对象,使用丹麦B&K公司生产的PULSE振动测试分析仪,使发动机以1500r/min的转速运转,并通过人为改变不同的缸壁间隙工况下采集振动信号,利用小波包变换提取缸壁表面振动信号的能量特征值,对提取出来的特征值,进行Elman人工神经网络的训练,其中将采集的40组数据作为训练样本,剩余的20组数据作为测试样本,从而识别出发动机缸壁的缸壁间隙,最后得到以下结论:基于小波包和Elman人工神经网络训练相结合的方法,对发动机缸壁间隙进行判断识别,通过实验的正确率验证了方法的可行性。

This paper mainly studies the DA462 - type engine, by theoretical analysis of selected Denmark B ＆ K company PULSE vibration test analyzer. In the circumstance of engine operating at the rotational speed of 1500r/min, vibration signal under different cylinder liner clearance is collected, a test bench of the engine cylinder liner clearance vibration test system is successfully built. Energy eigenvalue of the cylinder blook sur- face vibration signal is extracted using wavelet packet transform, some of which is used to train the Elman arti- ficial neural network. 40 sets of data are used as the training sample, and the other 20 sets of data is used as a test sample to identify the engine cylinder wall gap. We draw the following conclusions ： By combining inter- val wavelet packet and Elman artificial neural network training, we can identify the engine cylinder wall clear- ance, and prove the feasibility with the correct rate of the experiment.