Towards automated knowledge acquisition for process plant diagnosis

There exists a need for diagnostic monitoring systems to ensure safer operation of process plants. Although acquisition of plant knowledge for such systems would prove expensive and time-consuming using traditional techniques, we feel this could be significantly reduced by extracting the information directly from the data held in the CAD models used in plant design. Any additional knowledge requirement could be obtained using conventional expert/knowledge engineering techniques. Not only could this information be utilised for diagnosis, but, by use of object-oriented techniques, the same knowledge could be used to perform functions including the provision of an intelligent CAI system for operator training. The goal of our project is to investigate the link between the design model and the knowledge base, proving that the technique is feasible, and evaluating its worth. We have constructed a simple demonstration system to illustrate the automated knowledge acquisition technique, and this is able to build the plant-specific portion of a frame-based knowledge base from computer-aided process design (CAPD) files, for use in a diagnostic fault-finding system.