Detecting Failure Precursors in BGA Solder Joints Using Prognostic Algorithms

A failure precursor is a measurable parameter which provides an indication of impending failure. By detecting changes in these precursors, deviation of the system from its healthy state can be detected early and unanticipated failures can be avoided. The case study illustrates the use of Multivariate State Estimation Technique (MSET) and Sequential Probability Ratio Test (SPRT) to detect the onset of failure in Ball Grid Arrays (BGA) subjected to Accelerated Temperature Cycling (ATC) tests. The residuals from MSET show a shift in their distribution as they degrade; SPRT detects the anomalous behavior (degradation) in the interconnects. The time to detection of anomalies was found to be earlier than the time to failure of the BGA interconnects, providing a prognostic distance. Hence the change in resistance of the BGA can be used as a precursor to failure.