Data Logging and Reconstruction of Discrete-event System Behavior

摘要

In case of malfunctions or accidents related to an infrastructural system, it is useful to reconstruct and analyze the behavior that led to such an undesired situation. Understanding the behavior can help in improving the plant and the supervisory controller such that this situation is not encountered again. Many computer-controller mechanical systems use programmable logic controllers (PLCs) to implement the supervisory controller and to collect data from the system. Currently, incident analysis for PLCs often consists of plotting actuator and sensor signals to reconstruct and analyze the behavior. This way of analyzing is laborious and difficult to interpret for engineers not familiar with the system. In this paper, a different behavioral reconstruction and analysis method is proposed. In this method, models developed during the design of the supervisory controller are reused. From the collected data, a finite-state automaton is constructed. This automaton can be used for behavioral reconstruction via simulation, which is simpler and more intuitive. Moreover, by comparing the logged behavior with the behavior defined in the available models, faults can be identified. As a proof of concept, the behavior of a real movable bridge has been logged from a PLC, reconstructed, simulated, and analyzed.