Digraph-based models for automated HAZOP analysis

Hazard and operability (HAZOP) analysis is a systematic hazard identification procedure for predicting the various adverse consequences that might arise in a chemical plant due to abnormal causes. HAZOP analysis is typically performed by a group of experts by systematically considering all possible deviations from normal behavior of a process plant. This is a labor- and knowledge-intensive problem that can benefit from automation. Towards that goal, a framework using digraph-based models is described in this paper. The HAZOP-digraph models of the process units are developed in a process independent manner so that they can be used in a wide variety of process flowsheets. Based on this framework, an expert system for performing HAZOP analysis has been implemented in an object-oriented architecture using the expert system shell G2. A graphical HAZOP-Digraph model building tool is provided to let the user develop new models or add more knowledge to the existing models of the process units. The performance of this system was evaluated on an industrial HAZOP case study of a sour water stripping plant. The salient features of the expert system and its performance for this case study are described.