Abstract: The traditional statistical analysis-based reliability demonstration testing (RDT) requires a large number of samples and long testing time, failing to satisfy the development demand for short cycle and low cost of electronic products nowadays. We propose a system-level reliability accelerated demonstration testing (RADT) program for multi-failure mode products under multiple stresses by combining mathematical statistics and physics of failure (PoF). Based on the principle of allocating reliability target to each environment stress, products are decomposed at four levels, namely function, structure, mechanism, and stress, and the key function, weakness, dominant failure mode, and sensitive stress can be identified, respectively. After level-by-level allocations, the RADT plan and test profile can be established by connecting environmental testing and reliability testing. Firstly, system reliability is apportioned to each environmental stress based on fuzzy logic and analytic hierarchy process. Secondly, the stress-strength interference model is used to deduce the amplification coefficient for each stress. The overall acceleration factor can be derived by the principle of constant acceleration factor, and then the RADT scheme can be designed based on a statistical plan. A case example is presented to illustrate the effectiveness of the proposed approach in comparison with programs of GJB899A and IEC62506.

Key words: accelerated testing, reliability demonstration testing, reliability allocation, fuzzy logic, analytic hierarchy process, physics of failure