Industrials systems, including chemical industries, can be exposed to undesired events that may cause terrible accidents. These accidents must be controlled and reduced. To this end, numerous risk analysis management approaches have been aimed at reducing the risks to a tolerable level to avoid the catastrophic accident. This reduction is achieved by implementing several layers of protection, including organizational and technical barriers, this later known as safety-instrumented systems (SIS). The main objective assigned to a SIS is the detection of dangerous situations and implementation of a set of reactions necessary at a specific time to ensure that the equipment is under control. This function is typically name the safety-instrumented function and is characterized by a safety integrity level (SIL). A SIL is defined as a measure of the confidence to perform the safety function. This paper deals with the uncertainties of SIS using one of several robust probabilistic methods from a group of Bayesian networks (BN). A case study of an operational heater is used to illustrate the application. The results are then compared with the risk tolerance criteria and the safety of the improved process by updating the BN model.