Dealing with IC package material and design uncertainties using FUZZY finite elements

Abstract

In this work, a new method is proposed using Fuzzy finite element modelling (FFEM) which is a non-probabilistic approach to deal with uncertainties. This method requires one order lower number of simulations than the more commonly used Monte Carlo approach to predict the distribution on the output variables. The method is demonstrated on a thermo-mechanical stress case study with overmould material properties as input variables (uncertainties) and the package warpage at reflow temperature and the die stress at minimum temperature as output variables. For three well-chosen input uncertainties, the distribution of the output variables are predicted using the FFEM method, based on a parameter set of only 15 simulations. This distribution is verified with 1000 simulations based on the Monte-Carlo probabilistic approach.