SIL assessment of in-service safety instrumented systems in the chemical industry based on FBN-LOPA

Abstract

The Safety Instrumentation System (SIS) is a crucial safety device widely used in process industries. Its safety performance is measured by Safety Integrity Levels (SIL). However, for in-service SISs, traditional SIL methods face limitations in dynamic analysis and grading due to hardware degradation from prolonged operation. Therefore, this paper presents a Fuzzy Bayesian Network & Layers of Protection Analysis (FBN-LOPA) model to consider the operation of in-service equipment and realize the dynamic update failure probability to achieve dynamic SIL grading. The Bow-tie (BT) model is utilized to construct accident and fault trees, enabling a comprehensive analysis of failure modes in in-service SIS. Subsequently, the BT is mapped to a Bayesian Network (BN), allowing for dynamic analysis of accident scenarios through risk probability updates. To address the issue of incomplete and uncertain data in in-service SIS, the FBN approach integrates fuzzy logic with Bayesian inference, converting linguistic variables into probabilistic values. Finally, a SIL grading and verification of the in-service SIS was calculated with LOPA. Taking the Tennessee Eastman process (TEP) as an example, the proposed method reduces error by 14.6 % compared to the traditional HAZOP-LOPA method and dynamically updates the SIS failure probability, enabling more accurate and reliable SIL determination.