Fluid property reasoning in knowledge-based hazard identification: Dissertation

The study of serious accidents, which have occurred in the chemical process industry in recent times, highlights the need to understand fluid property related phenomena and the interactions between chemicals under abnormal process conditions or with abnormal fluid compositions. Consideration of these issues should be common practice in professional safety analysis work, and computer programs designed to support this work have to be able to deal with them. The purpose of Hazard and Operability (HAZOP) study is to identify all possible deviations from the way a plant design is intended to be operated and all hazards associated with these deviations. Due to its systematic nature, the method is a good candidate for automation. Several research groups have developed embryonic knowledge-based HAZOP systems. However, no automated hazard identification features are included in current commercial software packages supporting HAZOP. The main problem of knowledge-based HAZOP systems is their poor performance in relation to the correctness and completeness of the resulting HAZOP study. This thesis describes a novel methodology for fluid property reasoning in connection to knowledge-based HAZOP. Building on the earlier achievements of Loughborough University (LU) and Technical Research Centre of Finland (VTT) researchers, the methodology enables knowledge-based hazard identification programs to make a more intelligent assessment of the potential hazards and their causes. In the first phase of the study, a rule-based fluid property and reaction property reasoning system was created for use in the HAZOPTOOL program. In the second phase, the LU fault propagation reasoning methodology implemented in the AutoHAZID HAZOP emulation program was extended with fluid property reasoning capabilities. AutoHAZID was subjected to extensive evaluation which consisted of an evaluation workshop, a fluid model oriented study of the workshop results, and comparative testing based on a set of test cases. It was shown that it is possible and beneficial to extend knowledge-based HAZOP with a capability to reason about fluid properties and interactions. A framework for such a system is presented in this thesis together with some ideas for future work. Based on the results of the work reported here, it is recommended that fluid property reasoning is taken into use in any application of knowledge-based hazard identification.