A similarity based prognostics approach for real time health management of electronics using impedance analysis and SVM regression

Abstract

In electronic assemblies, interconnects provide connections between electrical components both mechanically and electrically. During their use conditions, interconnects encounter multiple stress conditions such as temperature extremes, corrosion, and vibration. When these stress conditions are applied repeatedly, interconnects may lose their functionality, and the assemblies may eventually fail during the operation. As the consequence of electronics failure becomes critical, the need for reliable interconnects will grow.This paper demonstrates a prognostic approach for predicting remaining lifetime of solder joints in real-time. During stress testing, impedance has been reported to respond to early stages of solder joint degradation in the form of non-linear increases. These early responses were trended using support vector machine and compared with real-time impedance measurements for solder joint remaining life prediction. To evaluate the prediction performance, the impedance of a degrading solder joint was collected during mechanical fatigue testing, and the proposed approach was applied in real-time. During fatigue testing, the proposed approach successfully predicted solder joint failures over multiple trials. These results imply this prognostic approach can be an effective means for real-time health management of electronics and can be applied to systems whose performance variable demonstrates gradual changes prior to their end of life.