Towards prognostics and health monitoring: The potential of fault detection by piezoresistive silicon stress sensor.

A piezoresistive silicon based stress sensor has been demonstrated successfully as an effective tool to monitor the stresses inside electronic packages during various production processes. More recently, the sensor has been evaluated as a sensor for Prognostics and Health Monitoring (PHM) systems. This paper presents a systematic approach that evaluates its performance from the perspective of failure mode detection. A detailed Finite Element method (FEM) model of existing test vehicles is created. The test vehicle consists of six DPAK (Discrete Package) power packages and three stress sensors. The results of simulation are verified by the signals obtained from the stress sensor as well as the supplementary warpage measurements. After inserting various failure modes into the model, statistical pattern recognition algorithms are implemented for fault detection and classification. The proposed technique can identify detectable failures during reliability testing by utilizing the database of stress sensor responses for healthy and unhealthy state. Thus, the results establish a baseline for the applicability of the piezoresistive stress sensor for an on-line monitoring PHM methodology.