Abstract
The subject for industrial safety becomes more and more emphatic with the progress of industry. It is urgent for industry to assess the safety and figure out the weakness of the system. Conventional probabilistic risk assessment (PRA) bases on probabilistic, statistic and reliability theory evaluate the relative importance of influential components and human behaviors using fault tree and event tree. The criteria for system design, operation and maintenance refer to the conclusions of PRA. PRA have applied for many years, and users approve its great functions, however, there are still some disadvantages of PRA, which conflict with real-life situation. The purpose of this paper is to modify these disadvantages utilizing Neuro-Network and Fuzzy theory. This paper applies the characteristic of fuzzy theory to deal with indefinite circumstance, transfers the binary conception in traditional PRA that components only have yes (normal) and No (failure) two discrete states of condition into membership conception in fuzzy theory that components have continuous states of condition. The idea of risk should combine simultaneously the chance of abnormal event to happened and its consequence if it does happen. It is very difficult to estimate the consequence of occurred event, and usually depend on the subjective judgement of expert in the process of estimation. This paper using fuzzy inference system to shift numerical variable into linguistic variable so that the experience and knowledge of operator can be implanted into the process of assessment easily. Human behaviors are uncertainty and unpredictable, and human reliability plays an important role in system reliability, therefore, it can not be ignored in system risk assessment. This paper employ theory and experimental data of HCR (Human Cognitive Reliability) model, combine with neuro-network and fuzzy applicant technique, to builds up a human reliability evaluation system, and expert’s opinions have implanted into this system. It is more simple and acceptable to assess, and the result can reflect the real-life situation by using this evaluation system.