WHAT IS PRA?

Probabilistic risk assessment (PRA) is a systematic and comprehensive methodology used to evaluate risks associated with complex engineered systems and facilities. In a PRA, risk is characterized by two quantities:

• the magnitude (severity) of the possible adverse consequence(s), and
• the likelihood (probability) of occurrence of each consequence.

PRA is now commonly used as a tool for assessing safety and identifying potential safety improvements in a wide variety of industries including aerospace and nuclear power.

Fundamentally, a Probabilistic Risk Assessment attempts to answer three basic questions:

• What can go wrong?
• What are the consequences?
• How likely are these consequences to occur?

THE PRA PROCESS

Generally, a PRA is manifested as an integrated mathematical logic model of the facility, subsystems, and associated human actions. The PRA models the various plant responses to an event that challenges plant operation. The plant response paths are called accident sequences. A challenge to plant operation is called an initiating event.

There are numerous accident sequences for a given initiating event. The various accident sequences result from whether plant systems operate properly or fail and what actions operators take. Some accident sequences will result in a safe recovery and some will result in system damage. For a nuclear plant this may result in damage to the reactor core with a potential for subsequent release of radiation. For many applications, the consequences are typically characterized as the discrete set, no damage, core damage, or significant release.

The accident sequences are graphically represented with event trees. Each event in the event tree (called a top event) generally depicts either a function or a system that is needed to respond to the initiating event. An analysis is performed for each top event in the event tree. This analysis is usually represented graphically with a logic model called a fault tree which is used to identify the combinations of equipment failures that lead to the function or system failure.

The quantification of the logic model relies upon actuarial data on the frequency of plant challenges and the probability of equipment failure, supplemented by expert judgment when data is unavailable. The probabilities of human failures are quantified using a systematic evaluation process that is based on the key performance shaping factors that have been shown to influence human reliability.

When the logic model is quantified it results in an assessment of the frequencies of the various consequences, together with a breakdown of which accident sequences are the most significant contributors. The total risk is characterized by the sum of the frequencies of the sequences that lead to the undesired consequences. Since the PRA model is built from the bottom up, the individual contribution from specific initiating events, equipment and human failures can be identified and used to guide plant safety improvements.