Bayesian Inference for the Calibration of Cohesive Zone Models of Dovetail Specimens From Laminated Composite Fan Blade

Abstract

Composite fan blades are the preferred alternative for the fan stage of most advanced high bypass ratio turbofan engines. The dovetail part bears a significant centrifugal load, and its ability to safely bear this load is one of the key points of the multilevel “test pyramid” approach of compliance demonstration. Debonding between adjacent layers is the main damage mode of laminated composite fan blades. However, there is difficulty in measuring the as-manufactured interlaminar mechanical properties used in finite element models. In this study, tensile loading was applied to simulate the interacting centrifugal force and capture mixed-mode damage evolution. Structural responses and material damages were calibrated with measured tensile loads through Bayesian inversion, where interface and contact elements with distinct bilinear behavior were selected. Posterior probability distributions of maximum interface tractions and contact stresses were solved using Markov chain Monte Carlo (MCMC) sampler. Results indicated that the two bilinear cohesive material models had a capacity of predicting empirical means of longitudinal reaction forces as that in test considering additional discrepancy term (0.035 kN and 0.96 kN respectively), while they made an significant impact on the prediction of tensile load history especially when two delamination cracks initiated and propagated. Interface elements provided a higher matching quality in predicting loading history and capturing damage mechanism in association with in-plane progressive damage analysis. This calibrated parameter set could be functioned as benchmark in numerically determining the ultimate tensile load of dovetail elements and reducing the necessary number of physical tests at elemental length level.