Abstract :
A hybrid method for systems analysis based on three generally available sources of system knowledge is presented. These three sources are a model of the system, human experience or expertise about the system, and recorded data from the system, and the method combines these to accomplish the required systems analysis task. In part, this makes explicit the often implicit combined use of model-based knowledge, expert knowledge, and recorded data. The framework for the method is a system model in the form of a directed graph, where the nodes represent processes and the arcs represent material or information flow. This model automatically identifies system parameters that may be causally related. Expertise is then applied to suggest the form of the relationships between these parameters. Analysis of monitored parameter data, thus directed by the model and the expertise, provides the final system characterisation. After this, the required objective is addressed. A concrete example of the method is given with respect to a simulated rubber seal manufacturing process. An important theme of the method is its automatic, easy to use, and highly visual approach
Keywords :
directed graphs; inference mechanisms; knowledge based systems; knowledge representation; learning (artificial intelligence); statistical analysis; systems analysis; causally related parameters; directed graph; expert knowledge; human experience; human expertise; hybrid method; information flow; material flow; model-based knowledge; recorded data; rubber seal manufacturing process; systems analysis;
