Application Of Neural Network On PLC-based Automation Systems For Better Fault Tolerance And Error Detection

Abstract

Neural networks have a wide range of applications such as building complex equations using the input and output characteristics of functions, predictions of outputs, error detections, monitoring complex systems, etc. Neural network’s capabilities of monitoring the system, error detection, and predictions merged with Programmable Logic Controllers (PLC) can improve the fault tolerance and error detections in automation systems. While the PLC program is being tested in the simulated environment before it is implemented in the automation system, the values of PLC’s I/O ports, timers and critical variables during the execution of the program can be used to train a neural network and prepare it to monitor the system. Execution of the trained neural network in parallel with the PLC’s execution where the inputs and the outputs to the PLC are also supplied to the trained neural network, adds an artificial intelligence inspired system monitor. A neural network based system monitor learns the characteristics of the automation system using PLC’s port values and internal variables during the training. A successfully trained neural network can detect a malfunction or abnormal behavior in the automation system when the outputs to the automation system generated by the PLC and the outputs generated by the neural network are compared. The abnormal behavior of an automated system could have been caused by intrusions in which the PLC code has been altered by external entities, hardware faults, malfunctions on PLC’s I/O ports, mishandling of the system by the operators, etc. Addition of AI-based monitor to the automation system provides an additional layer of security and helps the system run efficiently since neural network’s prediction capability can alert the operators if abnormal behavior in the system starts to take place before it is too late to recover.