SAFETY VERIFICATION USING A HYBRID KNOWLEDGE-BASED MATHEMATICAL-PROGRAMMING FRAMEWORK

Hazard and operability analysis (HAZOP) is widely used to perform haza rds analysis of chemical plants. It is labor-and knowledge-intensive a nd could benefit from automation. Toward that goal, a knowledge-based framework for automating HAZOP analysis (HAZOPExpert) was proposed. Re cently, Dimitriadis et al. proposed a quantitative model-based approac h that uses a dynamic model of the plant and a description of process disturbances and parameters for hazard evaluation. These two different approaches have their own merits and demerits. The qualitative analys is performed by HAZOPExpert is thorough and computationally efficient, but can lead to ambiguous conclusions. The quantitative approach can perform an exact analysis without ambiguities, but a complete analysis can be computationally prohibitive. Thus, these two frameworks appear to complement each other. This article presents an integrated approac h for hazard identification and evaluation which overcomes the shortco mings of purely qualitative and quantitative methods. In the integrate d framework, the overall features of a particular hazardous scenario a re extracted by inexpensive qualitative analyses. If necessary, a deta iled quantitative analysis is then performed and that too only on thos e parts of the plant identified by the qualitative analysis as hazardo us. The results of this framework are compared to those of purely qual itative reasoning using an industrial case study.