A fault diagnosis system-PLC design for system reliability improvement

PLAT: An Automated Fault and Behavioural Anomaly Detection Tool for PLC Controlled Manufacturing Systems

“…If a fault occurs in the control process, it completely alters the sequence of status change events of the PLC signals (in other words, the signal-state transition sequences of the devices and groups are completely modified) [10–14]. So, if PLAT finds any signal-state of a device or group that is not one-step reachable from its immediate previous signal-state (according to the corresponding nominal signal-state I/O model), then we can easily conclude that a fault has taken place in the system.…”

“…This gives us speed gain during fault and behavioural anomaly searching phase (by significantly cutting down the size of the model). In PLAT, a fault (or an anomaly that has significant impact on system operation) cannot remain unidentified because, in that case, the system fails to produce the exact same sequence of transitions (with almost the same transition time, transition, and transition time probabilities) as in the nominal signal-state I/O models [11, 12, 14] [we should mention that most of the existing automaton (see for instance: [9–14]) and event-sequence based approaches (see, e.g., [2, 3]) also provide the same fault detection accuracy rate, as they use the same fault detection principle (i.e., detect the faults based on whether there is any deviation in the sequence of signal status change events); anyhow, they do not solve the complete problem of real time fault and behavioural anomaly detection in large manufacturing systems]. So, the above stated model-size reduction is done without losing any useful information related to faults or behavioural anomalies associated with control process.…”

“…Similarly, if any subsystem or group stops producing the log data records for a very long period of time, then also we can conclude that a fault has occurred in the control process (identified by using the timer counters, one for each group). Please note that a fault cannot remain unidentified in this process because a system cannot generate the exact same transition sequences of the devices and groups (as in the index hash table) after the occurrence of a fault (in other words, a fault will eventually be detected since the observed faulty behaviour will eventually be distinguishable from the fault-free behaviour) [11, 12, 14]. If the transition time for a particular query transition lies below a certain threshold (user-defined, usually very small) of its corresponding maximum transition time, then we conclude that it is a “correct” transition.…”

“…In these types of approaches, the nominal manufacturing process behaviour is represented by a NDAAO type automaton and the faults are detected based on whether the observed system behaviour is following the transition behaviour of the NDAAO type automaton. In [14, 15], a (deterministic) Finite State Machine (FSM) based approach was proposed for fault detection. In another paper [16], the P-invariant of Petri Nets was applied to discover the sequence faults and the exclusive logic functions were applied to detect the sensor and actuator faults.…”

Design, simulation and experimental investigation of a novel reconfigurable assembly fixture for press brakes

“…The programmable logic controller (PLC) is controlled by the reconfiguration model which is installed on a computer. A fault diagnosis system will be used with the PLC in order to improve its reliability[61,62]. The picture of the experimental set-up is presented in Figs.…”