| Three thermal fatigue failure models for wires and wire bonds in microelectronic packages have been analyzed in detail. The parameters which most greatly influence model results were identified, and values for the magnitude of their variability determined. A simulation was created to probabilistically evaluate the three models. A generic microelectronic package with realistic dimensions, material properties, and parameters was identified and used as the basis for model evaluation. Three different statistical distributions were used to describe parameter variability.; The ability to refine model results based on actual sampling or testing data was also examined. The accuracy and responsiveness of parameter updating using a frequentist approach was compared to that when using a Bayesian approach. Analyses were conducted to determine if only a small number of parameters could be modeled probabilistically with little loss in overall accuracy. Analyses were also performed which showed the impact of modeling the failure mechanism models dependently (i.e., using the same values for common parameters for a given simulation trial) versus independently. Additionally, the impact of model coefficient uncertainty was quantified. Overall, the advantages and difficulties associated with probabilistically modeling the three failure mechanisms were addressed. The approach developed can be extended to probabilistically model and evaluate other failure mechanism models as well. |
