Domenic Di Francesco

Student Name

Domenic Di Francesco

Research Title

Application of Bayesian decision analysis to structural integrity management

Keywords

Bayesian modelling, dependency, fatigue crack growth, Markov chain Monte Carlo (MCMC) sampling, information theory

Sponsor

Lloyd’s Register Foundation

Affiliated University

University of Surrey

Supervisors

Dr Ujjwal Bharadwaj (TWI Ltd.), Prof Marios Chryssanthopoulos, Prof Michael Havbro Faber and Dr Boulent Imam (University of Surrey)

Start Date

October 2017

Project Outline

My research to-date focuses on Bayesian methods of quantifying multi-variate uncertainty (and dependency) in fatigue crack growth models. Model evaluation criteria demonstrated an improved out-of-sample predictive performance for Bayesian models (when compared with existing industrial guidance), with the most accurate predictions resulting from multi-level (hierarchical) models which account for variation between constituent test datasets and partially pooled information.

This project has also developed into a multi-stage (pre-posterior) decision framework for the quantification of expected value of inspection activities - conditional on mathematical characterisation of inspection precision, bias, reliability and risk - and a prior model of structural condition.

Future work will include the application of partial pooling models to infer problems in structural condition at locations remote to where an inspection has been completed.

Publications

Quantification of the value of inspecting for fatigue cracks, conference abstract & presentation, NSIRC 2019

Bayesian fatigue modelling: application of Markov chain Monte Carlo sampling to account for variability and dependency, conference paper & presentation, international probabilistic workshop (17th IPW) 2019

Inspection planning of hazardous locations using a value of information assessment, conference paper & presentation, international forum on engineering decision making (11th IFED) 2019

Thesis

Analysis of dependencies in Bayesian decision theoretic approaches to structural integrity management