Dispersion heterogeneous recurrence analysis and its use on fault detection

Recurrence plot is an effective tool for portraying system dynamics. However, dealing with distance matrices through the Heaviside function which is difficult to determine the threshold may lead to the loss of much important information, such as information about transitions between states. Therefore, in this paper, we propose a novel dispersion heterogeneous recurrence analysis for complex systems to explore their intrinsic characteristics and structure. The use of dispersion patterns in symbolic dynamics can retain valuable information better than the original defined recurrence plots and avoid the challenge of choosing a threshold. Moreover, we use the iterated function system to provide a visual display of the transition information between patterns. Finally, attention entropy is used to develop a dispersion heterogeneous recurrence quantification analysis. Experimental results show that the method is able to detect the changes of system characteristics with parameters. Also, it can be combined with clustering and classification algorithms for fault detection of railway vehicle systems. •We propose a novel dispersion heterogeneous recurrence analysis for complex system.•Dispersion patterns can retain valuable information and avoid choosing a threshold.•We use Iterated Function System to provide a visual display of pattern transitions.•The method can be combined with classification algorithms for fault detection.