Event - Driven , Asynchronous Control and Monitoring by Aaron

This thesis presents and explores an atypical method of designing a modern (computer driven) digital monitoring and control system. There are two tenets which distinguish this method from others. The first is the realization that, from the perspective of a computer driving the system, the monitoring processes and control processes are very similar: they both require the computer to send a request to an external instrument, and then lie in wait for a response. The second is that there is a natural division between the user interface(s) and the instrument driver(s) that can be exploited to lighten the load on any one computer. From these tenets derive the "concentrator" and the "observer". A concentrator is a conceptual entity that is in charge of driving external instruments (through autonomous processes) to observe important measurements or control important settings within the system, convert them to relevant values, and then buffer the information. An observer is the graphicals, the user interface, and in charge of any special calculations. It is from here that the buffered values are interpreted into expert decisions, either by an algorithm or an operator. Concentrators and observers are, by definition, independent and can therefore be run on separate machines (arbitrary numbers of each) communicating over a network. The minimal configuration, however, is one concentrator and one observer; the concentrator cannot run without first having a parameter file downloaded from an observer, and the observer would be useless without a concentrator. The beauty of this system comes from its flexibility and expandability. A new process can be added to a concentrator within minutes, and likewise its values' interpretation on the observer. These processes need not rely on timed loops to generate actions within the system. Every process is event-driven, generating an event (and triggering a process to run or data to flow) based on any imaginable occurrence; a mouse-click, or an external trigger, or a separate timed loop if required. Furthermore, there are no bottlenecks in the system due to one parameter having a slower loop than others; all processes are independent and therefore asynchronous, reporting data and setting values when needed, not just when available. Thesis Supervisor: Chathan M. Cooke Title: Principal Research Engineer