Probabilistic Component Design Assessments in Plant Systems Design: a Case Study

An ASME Plant Systems Design (PSD-1) Standard is being developed to provide a new approach to design facilities with potential for significant hazards to the health and safety of the public, the worker and the environment. The standard includes requirements and guidance for design organizations to incorporate risk informed Probabilistic Design (PD) Methodologies with traditional deterministic design methods using reliability and availability targets. The PSD-1 standard is organized around a typical design cycle with evolving requirements for Initial, Preliminary and Detailed design. This paper presents a case study of the Detailed Design PD process which focuses in demonstrating that individual components meet their reliability and availability targets. The case study describes the activities taken to select components or regions where probabilistic assessments offer the most benefit, identify the failure modes for inclusion in the model, select mechanistic models and inputs and perform and evaluate the PD analysis. In this case a probabilistic assessment was carried out for a weld between a forging and a tube. The main failure modes and mechanisms of concern were fatigue crack growth of an initial defect leading to brittle failure or initiation of tearing. This was assessed using the Monte Carlo approach with variables such as the initial defect size and position, residual stress, crack growth rate and toughness sampled from distributions. To reduce calculation times and maximize the number of cases that could be run, a simplified approach was used to model the fatigue crack growth, which had been benchmarked against a small number of deterministic crack growth calculations. The results are presented in the form of a probability of failure occurring at different points in a thermal transient occurring in the final year of operation. The case study then considers the effect of a change in requirements which is common during a complex design process. In this case, the required life was increased, resulting in an increase of the failure probability by up to four orders of magnitude. It was identified that the growth calculations were overly conservative resulting in unrealistically large end of life defect sizes. This had previously not been a concern as the vast majority of final sizes in the baseline case were less than the size needed for failure to be predicted. As a result, the growth was therefore refined and the assessment was re-run to improve the result, returning it to a similar level to the initial baseline.