Study Of Reliability Modeling For Safety Instrumented System Based On Markov Model Method

While the rapid development of industry has brought great benefits to mankind, it also increased the probability of accidents. In order to make industrial production more secure and reliable, more and more Safety Instrumented System (SIS) was applied in different fields to protect personnel from hazards. In order to reduce the frequency of disaster and ensure safe plant operation, it is very necessary to make effective reliability model for Safety Instrumented System.At present, the reliability modeling for SIS are mainly based on Fault Tree Model, Simplified Formula method, etc. Under some assumptions, these two methods are easy to use, but these assumptions often result in calculation error and invalidity. Compared to these two models, Markov Model can bring high accuracy.1. In order to achieve the reliability analysis and SIL calculation based on Markov Model for SIS easily, we developed the software MarkovSIL under Matlab7.1 and Vb6.0 environments. This software can effectively make reliability models for SIS and calculate PFD, PFS, SIL and some other failure parameters. It can be used for the design and evaluation stages for SIS, with user-friendly, flexible-modeling and easy-to-handle characters.2. To resolve the problems that the recycle hydrogen furnace unit has high risk of fire and explosions, we used MarkovSIL and Risk Graph techniques to redesign the system. By changing the equipment redundancy, TI (test intervals) and other parameters, we finally find the approach that meet the appropriate SIL level of SIS. By comparing the calculation results with Fault Tree model, we verify that Markov model produces more accurate solutions for many failure modes.3. We compared the independent and integration approaches for BPCS (Basic Process Control System) and SIS, and put forwarded the modeling method of 1oo2 Markov Model when the failure parameters of dual channels are different. Finally, we established the repairable Markov models in the imperfect detection situation by changing the model states and the transition matrix P. The result proved that Markov models are more flexible than other methods. This study can be used in future reliability analysis for SIS.