The legal requirement in the UK for the duty holder of a chemical process plant to demonstrate that risk is as low as reasonably practicable (ALARP) means that quantified risk assessments (QRAs) must be accurate and robust and that identified risks are adequately mitigated. Bayesian belief networks(BBN) is an emerging technique which can be used to determine the likelihood of an event in support of the QRA process. It is a statistical method involving estimating the probability distribution for a given hypothesis. The most interesting features which distinguish this QRA technique from all the others are: • it can analyse complex systems of any given number of variables and their dependability within a single analysis; • it can analyse parameters over a range of probability values for any given set of conditions, providing a better understanding in terms of sensitivity analysis; • it engages expert judgement and learning from previous events to update the probability distribution, thus improving QRA accuracy; and • it is not just restricted to fault analysis and can be used to support plant operational decision making using a quantified approach
