In this blog post titled: ‘Does Probabilistic Safety Analysis (PSA) need a reboot?’ CRA’s CEO, Jasbir Sidhu, shares his thoughts on this topic.
Probabilistic Safety Analysis (PSA) has been with us for a very long time especially with respect to its development and use in the nuclear industry.
Broadly speaking, the technique models the impact of an initiating event on a nuclear facility that could give rise to an undesirable consequence (such as a release of radiation). The technique relies on the development of event trees (representing the progression of an accident sequence) and fault trees (used to predict the reliability of the protection systems).
The model created represents the logical breakdown and interactions of the mechanical, electrical, control and instrumentation systems and human actions that ensure that the threat posed to the facility to maintain safe operations is mitigated. Data for each of the components making up the systems and human actions is sourced from operational experience, generic industry data, and human error quantification methodologies such as Human Error Assessment and Reduction Technique (HEART) and Nuclear Action Reliability Assessment (NARA).
Where there is a significant lack of data, say for a new component, system or process, expert engineering judgement is used. Eliciting data from experts can be tricky or problematic, thus to overcome some of the limitations of asking experts to provide their numerical opinions, techniques such as Structured Expert Judgement (SEJ) are deployed. At our 10th Risk and Safety Forum Dr. Ir. Tina Nane will be presenting a talk on SEJ.
The first applications of PSA concepts in the nuclear safety field were for gas-cooled reactors in the UK, in the 1960s. Criteria developed and produced by Farmer was crucial to this advancement happening. Over the years, the requirement to perform a PSA has been adopted by most, if not all, nuclear regulators around the world.
In the UK nuclear regulatory framework, a licensee of a nuclear facility has to provide a Safety Case. The Safety Case has two main legs – the deterministic safety assessment and the probabilistic safety assessment. The deterministic safety assessment tends to be qualitative and conservative, whereas the probabilistic safety assessment is numerical and largely best estimate. Dr Srinivas Golakoti from the Office for Nuclear Regulation (ONR) will be discussing how these two assessments can work in harmony from a regulatory perspective at this year’s Forum.
At the end of the day, PSA provides a numerical value of the likelihood of an undesirable event occurring, which is supported by a large amount of information that can provide numerous risk insights that could lead to improved design and operations for safety and increased output from the nuclear facility. This in turn has benefit for both society and the planet.
Today, many non-believers of the benefits of PSA suggest that:
*All you need is the deterministic safety assessment, PSA provides no added value.
*The models are overly complicated – you can’t see the wood from the trees.
*The assessment borders on mystical activity.
*Unnecessary cost – spend the money invested in a PSA on real changes on plant.
*It is only needed for regulatory compliance i.e. a tick box mentality.
Some of this criticism is probably valid. Some analysts can have a tendency to develop larger and more complex models as computing power increases. Poor communication to stakeholders of the complex risk picture that is unfolding does not aid matters either. Additionally, at times, we are not very good at explaining the benefits of the techniques to a wider non-analytical audience. The numbers we generate can confuse the recipient, leading to mistrust. Thus, careful communications are needed. (CRA’s 2018 Risk Forum, ‘My number, your number, our number. Managing expectations of all stakeholders explored the theme of what lies behind the numbers and touched upon this issue).
At CRA we aim to avoid this by ensuring the scope is understood on both sides from the beginning of the project. We recognise the need for communication, so we provide regular updates to inform of progress and to prevent deviation from scope or over-engineering.
Looking ahead, it is clear to me that PSA does need some change. Perhaps it can be rebranded to simply ‘Risk Modelling’. It certainly rolls off the tongue more easily than Probabilistic Safety Analysis. Also, the way in which we present our complex risk picture needs to be developed and communicated so that it can be accessed by the many rather than the few.
CRA has been developing risk tools to help managers and plant personnel to visualise which plant systems and components are most risk significant. Effective risk visualisation helps to easily communicate for which components it is cost effective to improve reliability. In the long term CRA plans to collaborate with EDF France R&D to introduce Risk Visualisation into the EDF PSA management tool.
Attending the Risk & Safety Forum is a fantastic opportunity to hear from industry experts, discuss latest developments and challenges and share best practice with peers.