Supervisor: Prof. Georges Kouroussis –

Co-supervisor: Prof. Olivier Verlinden

Abstract :

This thesis investigates the adhesive bonding of ceramic-bodied electronic components to space printed circuit boards (PCBs) subjected to launch vibrations.
A first contribution dealt with the identification of elastic and viscoelastic properties of the PCB and of a set of structural adhesives, viz., CV2946, EC2216, and Ablestik82, via static bending and vibrational tests, respectively. The data from static tests, consisted of structural responses of cantilever (uniaxial) and plate (biaxial) PCBs, while the vibrational test data encompassed the frequency response functions (FRFs) of a two-block adhesive assembly excited in shear/tension. The property identification procedure for the PCB/the adhesives relied on the correlation between measured and finite element (FE) simulated structural responses/analytical FRFs.
A second contribution covered the characterization of the mechanical resistance of adhesives in static and cyclic PCB bending. Static fracture responses of ceramic-PCB adhesive assemblies in tandem with PCB-only responses provided valuable elastic and failure limits of the tested adhesives. Furthermore, a FE modelling cohesive approach was designated (1) to extend outcomes of fracture tests to the identification of intrinsic cohesive properties of the tested adhesives, (2) to safely predict the adhesive stress, which contributed to develop an analytical curvature-based criterion of electronic adhesive assemblies, and (3) to evaluate the static damage in any location of the adhesive joint. A substantial effort was required to address the physical inconsistency of an existing bilinear cohesive model.
In the end, a vibrational test method was developed to excite a component at its resonance with displacement control. Applied to fatigue tests of uniaxial and biaxial ceramic-PCB adhesive assemblies, it provided first and second datasets of the PCB deflections versus cycles at the initiation of failure and cycles at failure, respectively. The first dataset was used to settle and calibrate a strain-based Basquin fatigue law, while the second dataset to examine the fatigue law relevance. The capability of predicting cycles to failure is promising compared to experiments.