Condition diagnosis with complex network-time series analysis

In this paper, we have introduced a novel method for condition diagnosis of complex systems in the chemical process industry with complex network based time series analysis. Firstly, by a computational method, the condition data from the complex system can be mapped into a network, which inherits the properties of condition data. Then, the topological properties of these complex networks are investigated, and the topological properties can be the features of condition. Finally, Dempster-Shafer (DS) evidence theory is applied to combining multiple features to find the underlying condition. A remarkable case study is provided to illustrate the method and to test its effectiveness. Results show that different condition data of system exhibit distinct and different topological properties; and the proposed methods can detect potential abnormal conditions from the normal condition effectively. This approach is markedly different from conventional methods, and can overcome the disadvantages in application of data-driven methods to condition diagnosis of complex system.