基于时延SDG和ICA的多工况过程故障预测方法  

A Fault Prognosis Method Based on Time-Delayed SDG and ICA for Multi-Mode Industrial Processes

作　　者：陆宁云[1] 王磊[1] 姜斌[1] 

机构地区：[1]南京航空航天大学自动化学院,江苏南京210016

出　　处：《控制工程》2011年第4期632-635,654,共5页Control Engineering of China

基　　金：国家自然科学基金(20806040);高等学校博士学科点专项基金(20070287047)

摘　　要：针对多工况生产过程,提出基于时延符号有向图(SDG)和独立成分分析(ICA)的在线故障预测方法。时延符号有向图描述了过程变量间信息传递的方向和时延大小;在有向图模型基础上,对过程数据进行信息同步校正;ICA方法应用于校正后的过程数据,使得基于ICA的性能监测方法具有良好的故障预测能力。在空分设备中的应用结果表明,该方法可在变工况下实现准确的氮塞故障预测。A fault prognosis method for multi-mode processes is proposed by integrating time-delayed signed digraph（SDG） and independent component analysis（ICA）.Time-delayed SDG model is developed to present the information direction and the time lags between process variables.According to the SDG model,the process data is preprocessed to highlight event information before ICA is applied for process monitoring and fault prognosis.The proposed method is applied to an air separation unit and achieves satisfying results in predicting the ＂nitrogen-block＂ fault when the unit is under operation mode changeovers.

关 键 词：多工况过程 故障预测 符号有向图 独立成分分析 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]