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

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.