Genetic Algorithm Optimization Model for Determining the Probability of Failure on Demand of the Safety Instrumented System

A more accurate determination for the Probability of Failure on Demand (PFD) of the Safety Instrumented System (SIS) contributes to more SIS realiability, thereby ensuring more safety and lower cost. IEC 61508 and ISA TR.84.02 provide the PFD detemination formulas. However, these formulas suffer from an uncertaity issue due to the inclusion of uncertainty sources, which, including high redundant systems architectures, cannot be assessed, have perfect proof test assumption, and are neglegted in partial stroke testing (PST) of impact on the system PFD. On the other hand, determining the values of PFD variables to achieve the target risk reduction involves daunting efforts and consumes time. This paper proposes a new approach for system PFD determination and PFD variables optimization that contributes to reduce the uncertainty problem. A higher redundant system can be assessed by generalizing the PFD formula into KooN architecture without neglecting the diagnostic coverage factor (DC) and common cause failures (CCF). In order to simulate the proof test effectiveness, the Proof Test Coverage (PTC) factor has been incorporated into the formula. Additionally, the system PFD value has been improved by incorporating PST for the final control element into the formula. The new developed formula is modelled using the Genetic Algorithm (GA) artificial technique. The GA model saves time and effort to examine system PFD and estimate near optimal values for PFD variables. The proposed model has been applicated on SIS design for crude oil test separator using MATLAB. The comparison between the proposed model and PFD formulas provided by IEC 61508 and ISA TR.84.02 showed that the proposed GA model can assess any system structure and simulate industrial reality. Furthermore, the cost and associated implementation testing activities are reduced.