AHA: a knowledge based system for automatic hazard identification in chemical plant by multimodel approach

摘要

AHA (automatic hazard analyzer), an expert system with new process knowledge models and inference algorithms for hazard analysis, is developed and tested. A multimodel approach is used to build better process models suited to chemical process. Knowledge representation models are composed of a unit knowledge base, an organizational knowledge base and a material knowledge base. Three hazard analysis algorithms (deviation, malfunction and accident analysis algorithm) are proposed. AHA is developed using expert system shell G2. The unit knowledge base is devised to model a process unit. It consists of a unit behavior model and a unit function model. In the unit knowledge base, a process unit is modeled in different terms of variable and function. This model represents physical hazards. The organizational knowledge base gives information about spatial arrangement of process units and streams. In a material knowledge base, material properties are considered according to NFPA code. This system performs hazard analysis in terms of both functional failure and variable deviation and thereby improves the quality of analysis and more possible accidents can be identified. The result of analysis provides a pathway leading to an accident, and, therefore, gives not only clear understanding of the accident, but useful information for hazard assessment. Using AHA, proposed methodology is applied to the feed section of an olefin dimerization plant, and performed better than traditional qualitative hazard analysis methods such as HAZOP study.