Modeling and analysis of linkage fluctuation for industrial process based on complex network theory

摘要

An industry process includes a variety of monitoring variables, and there is a strong relevance between the variables. Time series from different monitoring points can be used to obtain the feature of system's linkage fluctuation. Using the coarse-grained method, monitoring series has been changed into a sequence of characters. The sequence which is sliding by a certain length is converted into continuous modes of linkage fluctuation. The complex network of linkage fluctuation is composed of all the modes and the link edges between them. Transformation law for modes has been analyzed by degree distribution and structure entropy of complex networks theory. The matrix of structure entropy is used to reflect the operating state of the whole system. This method has been applied to analyze typical faults in monitoring data from large-scale coal chemical compressor units. It is indicated that the modes both in the normal operation and fault condition follow a power-law distribution. Transmission is mainly finished by few modes, but the structure entropy of networks have obvious differences. The results show that the ordered features of the linkage fluctuation between the process monitoring series can reflect the operation quality of the system, and the quantitative evaluation for the order state provides a method for monitoring the operation quality of the industrial process.