BGA solder strain prediction using an Artificial Neural Network regressor

Abstract

The increasing market demand for reliable electronics needs to go hand-in-hand with the development of predicting tools that help assessing the assembly lifetime in order to 1) Assess and quantify the main risks for failure and 2) Have confidence that the product will pass reliability tests. Here, we present a method to rapidly calculate the inelastic strain of solder joints (related to lifetime) in any PBGA component under arbitrary thermal cycling. The method relies on the application of an Artificial Neural Network (ANN) to fit the inelastic strain resulting from parametrized simulations with 15 variable parameters. The predicted strains show a standard deviation from the simulated strains of about 15%, which is acceptable for being used in a lifetime predicting tool.