Damage tolerance reliability analysis combining Kriging regression and support vector machine classification. (July 2019)

• Record Type: Journal Article
• Authors: Chocat, Rudy; Beaucaire, Paul; Debeugny, Loïc; Lefebvre, Jean-Pierre; Sainvitu, Caroline; Breitkopf, Piotr; Wyart, Eric
• Is Part Of: Engineering fracture mechanics. Volume 216(2019)
• Publication Date: 2019-07
• Subjects: Damage tolerance -- Fracture mechanics -- Reliability -- Kriging -- Support vector machine -- Subset simulation; Fracture mechanics -- Periodicals; Rupture, Mécanique de la -- Périodiques; Fracture mechanics; Periodicals; 620.112605
• DOI: 10.1016/j.engfracmech.2019.106514
• Languages: English
• ISSNs: 0013-7944

Abstract

Highlights: Hypotheses for Failure Assessment Diagram (FAD) damage tolerance lead to consider quantitative and qualitative outputs for the FAD margin. A surrogate based reliability damage tolerance dedicated method is proposed coupling regression and classification surrogates. Methodology based on Subset Simulation is used to assess the low probability (<10 −6 ). The number of damage tolerance model evaluations is reduced by several order of magnitude for low probability assessment. Adaptive design of experiment enrichment strategy is required to improve the quality of surrogates. Abstract: Damage tolerance analysis associates a Fracture Mechanical model with the Failure Assessment Diagram to define the state of a space engine component. The reliability analysis treats the variability of numerical models assessing the probability of failure within Linear Elastic Fracture Mechanics (LEFM) hypotheses. However, these models, while providing quantitative information in the safe domain, give only qualitative information for failed components. This work proposes an original methodology to combine Kriging regression and the Support Vector Machine classification along with transition criteria between both approaches. To accurately describe the limit state, we define a specific enrichment strategy. The efficiency of the proposed methodology is illustrated on reference test cases.