Building on HAZOP (hazard and operability) studies to produce knowledge-based systems

One of the major problems in constructing knowledge-based systems is so-called verification, i.e. ensuring the system´s knowledge is correct in some sense. This problem is felt particularly acutely in safety critical systems, where of course inclusion of incorrect knowledge may threaten life. The safety of a new chemical plant is ensured by an extensive review of its design, control and operability called a Hazard and Operability (HAZOP) Study. This means that design data which has been reviewed through the HAZOP study can be regarded as conforming with the safe operation of the plant. This paper discusses an automatic method of extracting knowledge from the results of the HAZOP study for use in knowledge-based systems. Although additional knowledge (for example more general knowledge of physics or fault diagnosis) may be required to build a complete application, the method in effect provides verification `for free´ of a substantial part of the knowledge base