›› 2015, Vol. 21 ›› Issue (第2期): 519-527.DOI: 10.13196/j.cims.2015.02.026
Previous Articles Next Articles
Online:
Published:
Supported by:
孙锴1,2,高建民1,高智勇1,高旭1,王昭1
基金资助:
Abstract: In view of the mass monitoring data set accumulated by online monitoring subsystem,a new method was proposed by transferring monitor data set into two-binary digital image named system fault-spectrum to recognize the plant-wide fault pattern with the inherent characteristics of the system.The monitor data set was separated into two groups of normal data and abnormal data,and the system fault-spectrum was drawn by coloring these two groups with black or white.By introducing the digital image processing technology,the most similar fault-spectrum in the typical fault pattern database was searched with traditional digital image similarity algorithm,which could realize the plant-wide fault pattern recognition.Two cases were used to verify the validation of the proposed algorithm.
Key words: fault-spectrum, multi-variate time series, process industry system, complex electromechanical system, online monitoring, nonlinear
摘要: 基于流程工业实时监控子系统积累的海量监控数据集,提出一种利用系统固有属性将系统监测数据集转换为二值数字图像,进而构造系统故障图谱的新方法。通过分析故障图谱,在企业层面上实现了故障模式识别。利用监测数据正常运行时的监测值范围,将监测数据分为正常数据和异常数据两类。对两类数据分别着以黑色或白色,画出系统故障图谱。引入数字图像处理技术,利用传统的数字图像相似度算法,在故障模式库中搜寻相近的故障图谱,实现系统级别的故障模式识别。结合两个应用实例,验证了算法的正确性和实用性。
关键词: 故障图谱, 多变量时序数据, 流程工业, 复杂机电系统, 在线监控, 非线性
CLC Number:
N945.17
TH86
孙锴,高建民,高智勇,高旭,王昭. 基于故障图谱的企业级故障模式识别方法[J]. 计算机集成制造系统, 2015, 21(第2期): 519-527.
0 / Recommend
Add to citation manager EndNote|Ris|BibTeX
URL: http://www.cims-journal.cn/EN/10.13196/j.cims.2015.02.026
http://www.cims-journal.cn/EN/Y2015/V21/I第2期/519