Solder interconnection failure time estimation based on the embedded precursor behaviour modelling

Abstract

This article presents a method for estimating solder interconnection failure times based on nonlinear parameter estimation techniques in conjunction with the Levenberg–Marquardt Algorithm (LMA). The LMA algorithm is applied to precursor measurements, from which failure time estimations are deduced using nonlinear estimation parameters. An electrical behavioural model is also proposed for precursors indicating weakening of solder interconnections. This model is divided into three sections: linear region, conversion region and nonlinear portion. Results obtained from experimental studies are presented and discussed in detail. In addition, an error estimation of the results is conducted to obtain knowledge of the most critical parameters in the estimation problem. The results suggest that the proposed method is capable of predicting solder interconnection failure times in advance, provided that the precursor describing electronic fatigue damage in interconnections obeys the exponential growth model during the wear-out period of the interconnection life-cycle. The obtained prognostic distances (PDs) varied between 15% and 38%, and failure times were predicted 15–25% before the actual failure occurred with an error rate of less than 10%. To show the generality of the proposed method, experimental results are presented for several samples with both lead and lead-free solders.