Prediction of reliability is essentially required for design for reliability, warranty periods, life cycle costs and maintenance with prior to the installation of components. Reliability prediction of electronics advances from an era of military standard books, Telcordia, PRISM, Bell Core etc. to the physics of failure approach.

Most of the limitations of constant failure methods are being encounter by the PoF approach due to its advanced methodology of finding failure mechanisms by root cause of failure analysis. But this approach has its own challenges as it requires detailed information on materials, process, technology and other specifications and at most of the times this data is confidential from the industries. On the other hand, probability and statistics methods provide quantitative data with reliability indices from testing by experimentation and by simulations.

In this paper, qualitative data from PoF approach and quantitative data from the statistical analysis is combined to form a modified physics of failure approach. This methodology overcomes some of the challenges faced by PoF approach as it involves detailed analysis of stress factors, data modeling and prediction. A decision support system is added to this approach to choose the best option from different failure data models, failure mechanisms, failure criteria and other factors