Discrete Event Process Modeling of Manufacturing Systems Using Sensor Graphs

Abstract: The design of control programs for manufacturing systems is becoming more and more complex as the demands for flexibility, efficiency and reconfigurability increase, due to changing consumer demands and fierce competition. There are methods stemming from academia that can assist the control programmers in facing these challenges. Some of those methods require a formal model of the system to be controlled. This thesis presents the language Sensor Graphs for modeling systems that are to be controlled by a PLC. The system under control, the process, is typically mechanical, containing equipment such as lifts, fixtures and robots, as well as mobile objects like pallets and manufacturing parts. The controller observes the state of the process through binary position sensors and influences it through binary control signals. The Sensor Graph model describes, as a discrete event system, how the physical objects activate and deactivate the sensors under the influence of control signals. The purpose of the suggested language is to make modeling easier and less fallible for the kind of systems described above, compared to existing general purpose discrete event languages. The language has a graphical syntax that often makes the models more compact and concise than models expressed using classical state machine- or Petri net-based languages. It does also support hierarchical and component-based modeling. It is shown how a closed-loop system can be formally verified based on a Sensor Graph process model and a discrete state equation representing the PLC program. It is also shown how the Sensor Graph process model can be translated into an observer that provides the controller with an estimate of the process' state. This state estimate is used for state feedback control and automated fault monitoring. The semantics of Sensor Graphs takes care of the details concerning the interaction between the discrete event process and the discrete-time controller, for instance the computation delay between the controller's input sampling and its output response, and the timing of the process in relation to the controller.