Perpustakaan Digital ITB

This thesis presents the development, validation, and uncertainty analysis of a finite element cohesive zone model representing the Mode I fracture behavior of an adhesively bonded double cantilever beam (DCB) joint. The adherends consist of unidirectional carbon/epoxy laminates bonded with a ductile epoxy adhesive (Araldite 2015), following the experimental configuration reported by de Moura et al. (2008). Cohesive behavior at the bonded interface was represented using an energy-based bilinear traction-separation law, implemented in Abaqus/Standard as a surface-based interaction between two directly contacting adherend parts, after an initial modeling attempt using discretely meshed cohesive elements was found to produce persistent convergence errors. Model convergence was established through a mesh refinement study across four interface mesh densities, guided by the estimated fracture process zone length, with the predicted fracture energy converging to within 0.3 percent between successive refinements. Validation against the reference load-displacement data revealed that the literature-reported adherend longitudinal modulus produced a substantially over-sti! elastic response; a calibrated modulus, approximately half the reference value, was adopted after diagnostic checks identified single-edge boundary condition application and the surface-based cohesive interaction as the most plausible mechanical contributors to this discrepancy. The calibrated model was subsequently assessed against three digitized reference load-displacement curves, showing good agreement in both elastic slope and peak load. Following validation, uncertainty in six cohesive zone parameters, representing an assumed hypothetical range of material variability rather than measured manufacturing data, was propagated through the model using a Latin Hypercube Sampling scheme with twenty representative samples, following a methodological framework informed by Crusenberry et al. (2023). A thirty percent input uncertainty propagated to a coe”cient of variation of only 6.52 percent in predicted peak load. Sensitivity screening using Pearson and Spearman correlation and standardized regression coe”cients identified the Mode I fracture energy as the dominant driver of this variability, explaining approximately 98 percent of the observed output variance, while cohesive elastic sti!ness and damage-initiation strength parameters showed statistically negligible influence. These findings provide practical guidance for prioritizing manufacturing quality control e!orts, specifically adhesive fracture energy and ductility, in the production of adhesively bonded joints.