Abstract
Pressure Reliving Devices (PRD) forms the last line of defenses in a Process Safety system. Their performance is often tested only during the scheduled testing opportunities. A review process of validating the adequacy of the testing frequency is deemed to be of paramount importance from a Process Safety and Integrity perspective. A set of 351 PRD test data from crude unit of three refineries were studied to develop a reliability based model using three different approaches.
Risk based approach using API 581, Normal Exponential Reliability equations and 2 Parameter Weibull Analysis (using industry standard software) were used to calculate the reliabilities of the set of sample PRDs with different testing frequencies and benchmarked against the desired level of reliability.
The testing data of 351 PRD test data were analyased through the three techniques and reliability values calculated. The sample refineries used frequencies of 2, 4 & 5 years and 99% desired reliability. With the sample failure data, the plant using 4 years frequency was be closest to the desired level of Reliability. 2 years frequency was found to be too conservative (over Inspection) while the plant with 5 year frequency showed less then desired level of PRD reliability, thereby compromising Process Safety.
The assessment demonstrated usefulness of normal Exponential equation based calculations, which is reasonably consistent with the complex calculations done through the API 581 or the Weibull Software generated reliability values. It was evident that the Normal Exponential Equation with simple failure rate as the only critical input, was adequately robust to carryout preliminary assessment of adequacy of the PRD testing frequencies. The size and homogeneity of the database is critical in increasing the accuracy of the assessments. This approach would assist scientific validation of PRD testing frequencies. Potential for a software based platform for PDR reliability assessment also is an opportunity.
