Fusion prognostics-based qualification of microelectronic devices

Abstract

The rapid evolution of electronic products has resulted in numerous choices for customers. This has made for intense competition between manufacturers to reduce costs and minimize the time to market for their products. One bottle-neck in getting products to market is the qualification process, which has traditionally been time-consuming and often inadequate to prevent failures in field. In particular, in the past decade, there have been significant numbers of microelectronic devices that have passed qualification tests but failed in the field. The resulting costs of these failures have been in the billions of dollars. Thus, there is a need to develop approaches to qualification methodologies that quicken the development time but also prevent product failures in the field. This paper discusses the current state of qualification practices in the electronics industry. Then, an alternative approach, called fusion prognostics, for qualification is presented that can make the process more efficient and cost-effective. This approach involves an in-situ qualification process that incorporates a fusion of machine learning techniques and physics-of-failure based prognostics. The machine learning techniques are used to monitor the degradation behavior during testing. On the other hand, the physics-of-failure techniques identify critical failure mechanisms and the acceleration factors.