Bayesian networks: application in safety instrumentation and risk reduction

The Safety Integrity Level (SIL) of a Safety Instrumented Function (SIF) depends on failures of the various components involved in performing the function. These failures depend on various factors and can be random hardware failures and/or systematic failures. Failure of a SIF need not necessarily result in a hazardous event when there are other layers of protection. Hence the residual risk probability that is left out after various layers of protection is of interest and it should be tolerable. In order to find the residual risk due to a hazard we need to know the demand rate of the hazard, the failure rates of various layers of protection and the factors which influence these failures. So the failure rates are not static and are dynamic, as various factors come into play during the lifecycle of the protection devices involved. In this paper the author proposes Bayesian belief networks to build the scenario based hazard probability model and uses that in the post-design phase to track the residual risk probability. An example is used to illustrate the application.