Optimising non-destructive examination of newbuilding ship hull structures by developing a data-centric risk and reliability framework based on fracture mechanics

  • Peyman Amirafshari

Student thesis: Doctoral Thesis


Ship structures are made of steel members that are joined with welds. Welded connections may contain various imperfections. These imperfections are inherent to this joining technology. Design rules and standards are based on the assumption that welds are made to good a workmanship level. Hence, a ship is inspected during construction to make sure it is reasonably defect-free. However, since 100% inspection coverage is not feasible, only partial inspection has been required by classification societies. Classification societies have developed rules, standards, and guidelines specifying the extent to which inspection should be performed. In this research, a review of rules and standards from classification bodies showed some limitations in current practices. One key limitation is that the rules favour a "one-size-fits-all" approach. In addition to that, a significant discrepancy exists between rules of different classification societies. In this thesis, an innovative framework is proposed, which combines a risk and reliability approach with a statistical sampling scheme achieving targeted and cost-effective inspections. The developed reliability model predicts the failure probability of the structure based on probabilistic fracture mechanics. Various uncertain variables influencing the predictive reliability model are identified, and their effects are considered. The data for two key variables, namely, defect statistics and material toughness are gathered and analysed using appropriate statistical analysis methods. A reliability code is developed based Convolution Integral (CI), which estimates the predictive reliability using the analysed data. Statistical sampling principles are then used to specify the number required NDT checkpoints to achieve a certain statistical confidence about the reliability of structure and the limits set by statistical process control (SPC). The framework allows for updating the predictive reliability estimation of the structure using the inspection findings by employing a Bayesian updating method. The applicability of the framework is clearly demonstrated in a case study structure.
Date of Award29 Jul 2019
Original languageEnglish
Awarding Institution
  • University Of Strathclyde
SponsorsUniversity of Strathclyde
SupervisorNigel Barltrop (Supervisor) & Selda Oterkus (Supervisor)

Cite this