基于油液原子光谱多维时间序列模型的机械磨损状态监测研究

提出了一种利用润滑油原子光谱分析技术对机械磨损状态进行监测的新方法。对磨合期润滑油原子光谱数据建立多维时间序列模型并视为标准模型,将新数据通过此模型后得到残差并选择残差方差阵元素作为新数据所属磨损状态的特征。然后,利用主成分分析法对高维特征进行降维,提取前三个主成分构成对应磨损状态的特征向量。最后,利用欧式距离度量对测试样本进行分类,达到了对机械磨损状态识别的目的。利用上述方法,通过对某型履带车辆发动机台架实验的光谱数据进行分析,对发动机磨损状态进行了有效识别,从而证明了所提方法的有效性。结果表明,将多维时间序列模型引入油液光谱分析技术,能够实现光谱信息的有效融合,提高机械磨损状态监测的准确性。

Research on monitoring mechanical wear state based on oil spectrum multi-dimensional time series model

A new method using oil atomic spectrometric analysis technology to monitor the mechanical wear state was proposed. Multi-dimensional time series model of oil atomic spectrometric data of running-in period was treated as the standard model. Residues remained after new data were processed by the standard model. The residues variance matrix was selected as the features of the corresponding wear state. Then, high dimensional feature vectors were reduced through the principal component analysis and the first three principal components were extracted to represent the wear state. Euclidean distance was computed for feature vectors to classify the testing samples. Thus, the mechanical wear state was identified correctly. The wear state of a specified track vehicle engine was effectively identified, which verified the validity of the proposed method. Experimental results showed that introducing the multi-dimensional time series model to oil spectrometric analysis can fuse the spectrum data and improve the accuracy of monitoring mechanical wear state.