Updating Elicited Expert Judgements using a Bayesian Network

This paper presents the details of a case study which involves analysing Fan Blade-Off (FBO) events by using a probabilistic approach. This approach is based on Bayesian inference and it focuses both on the structures of the priors as well as the likelihood functions which represent the two concepts that any Bayesian updating procedure relies on. In the first part, the theory which has been used for implementing the framework is explained, and this includes the Bayesian updating procedure itself as well as one method of obtaining priors: by eliciting expert judgements. The case study presented is related to analysing in service FBO events and uses expert judgements as prior distributions. In addition to that, it is shown how Bayesian inference plays a role in making the code of dynamic in light of new data, In other words, whenever an engine failure occurs due to a blade segment breaking off, the new data is then input into the FBO model, and this ultimately has the effect of updating the model so that it reflects both the initial prior judgements as well as the newly available data. The Bayesian updating procedure is iterative in the sense that as more data becomes available, the model is incrementally updated so that its depiction of reality becomes more and more accurate. Overall, it is believed that the framework described in this paper will allow the aero-engine designer to preform relevant and more detailed analysis on the fan subsystem during the preliminary design process.