SCADA stands for Supervisory Control and Data Acquisition. This is a system that monitors and controls a plant or equipment in industries. Its application is in a wide range of industries such as power, telecommunication, water distribution, waste control, energy, oil refineries etc. A SCADA system comprises of data transfer between a central host computer and several other units. Those units are either RTUs(Remote Terminal Units) or PLCs(Programmable Logic Controllers), and operator terminals. Many of the traditional SCADA systems utilized open loop control systems but most modern ones utilize closed loop systems where certain configurations are setup usually by a controller via a HMI(Human Machine Interface). The SCADA system maintains the actuators within certain parameters for relevant configured variables. Because of the large amount of data provided by the SCADA systems on an almost real-time basis usually via a GUI(Graphical User Interface) or back-end provisioned sensors and controllers, they are prime candidates for utilization of Al and ML techniques. Currently a lot of utilization of ML is done to identify alarms and even predictive analysis of the same. In addition, ML(Machine Learning) is utilized in the determination of cyber threats by analyzing incoming traffic patterns via IP and bandwidth for example in cases of DDOS(Distributed Denial of Service) attacks. If these large systems can use the data to progress to a higher level of self-autonomy, this may release the human operators to more meaningful and interesting tasks and not the mundane. This paper is interested in a study of how to use of data provided via a conveyor based SCADA system model to attempt self-error identification, self-correction, and self-healing functionalities using ML based control techniques. To this end, because of the steep costs in setting up a fully-fledged system, this paper will utilize modeling tools that will simulate parts of an industrial production line managed via a SCADA system. In addition, this system will test a limited scope section of the model via an industrial logic controller namely Siemens S7-1200 PLC interfacing with a DC(direct current) motor actuator and with an STM32F4 Nucleo Board as a controller.

SCADA Machine Learning Artificial Intelligence Model Predictive Control MPC NARMA-L2 Nonlinear Auto-regressive Moving Average-L2