复杂工业过程的关联波动网络建模与分析

Modeling and Analysis of Correlation Fluctuation Network for Complex Industrial Process

流程工业过程监测变量众多,过程关联性强,过程监测时间序列能够反映变量之间的关联波动规律.依据粗粒化方法将监测时间序列的关联波动状态转化为字符序列,序列按一定长度滑动形成一定长度的字符串的连续关联波动模态.将模态作为节点,模态之间的转化为边,构建监测时间序列之间的关联波动复杂网络,并运用复杂网络的相关理论对关联波动模态的变化规律进行分析.将其应用于煤化工大型压缩机组工作过程中典型故障监测数据分析,结果表明:系统正常运行和发生故障时监测序列关联波动过程的模态分布均具有幂律性,模态转换和演化主要发生在少数模态之间,但模态变化规律差异明显.结论不仅用于复杂工业过程的故障预警,同时也为探索多变量之间的关联波动规律提供了一种思路.

Process industry includes a variety of monitoring variables,there are strong relevance between variables,monitoring time series can reflect the relationship of correlation fluctuation between the variables.Using coarse graining process,monitoring time series is changed into a sequence of characters,the sequence sliding by a certain length is transformed into a continuous linkage fluctuation modes.The complex network of linkage fluctuation is composed of all modes and link edges between them.The variation law analysis for modes is analyzed by complex networks theory. This method is applied to analyze typical fault of monitoring data from large-scale coal chemical compressor units.It indicates that the modes from the normal operation and fault condition of the system have power-law distribution,transmission is finished mainly by few modes,but the laws have obvious difference.The results not only can be used for complex industrial process fault warning but also provides an idea for researching general law in different monitoring variables.