多尺度正交时序模型及其在网络化设备预维护中的应用  

Multiscale Orthogonal Time Series Model and Its Application to Predictive Maintenance for Networked Machine

作　　者：黄亦翔[1] 刘成良[1] 李彦明[1] 

机构地区：[1]上海交通大学机械与动力工程学院,上海200240

出　　处：《上海交通大学学报》2008年第7期1168-1172,共5页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金资助项目(50575145);国家高技术研究发展计划(863)项目(2006AA04Z432;2007AA04Z419)

摘　　要：针对设备智能维护领域中网络化趋势所引起的分布计算及传输问题,提出一种多尺度正交时序预测模型.利用正交小波变换的去相关特性,对各尺度上的分解系数序列分别建立相互独立的时序模型,以得到多尺度正交时间序列模型.将其作为预测模型应用于智能维护,通过实测数据验证了该模型的可行性.与单一时序建模方法相比,多尺度正交时间序列模型充分利用分布式特点,更适用于网络化设备的预维护.For the new problems of transmission and distributed-computation introduced by the trend of networking in the field of intelligent maintenance for equipments, a multiscale orthogonal time series model was proposed. Due to the orthogonal wavelets＇ decorrelation property, the series of wavelet coefficients at different scales can be obtained by ortbogonal wavelet decomposition. Then, by establishing the time series models at different scales independently, the muhiscale orthogonal model was derived. The multiscale orthogonal time series model can be applied as a prediction tool for trend estimation and its feasibility was validated by a real data set. Compared with the single time series model, the new method can achieve distributed computing and storing and adapt to various network environments. The multiscale orthogonal time series model is more suitable for the distributed intelligent maintenance.

关 键 词：预维护 小波变换 正交时序 多尺度模型 趋势估计 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置] TH165[自动化与计算机技术—控制科学与工程]