Statistical analysis of power cycling data

Power cycling is a common test method to assess the lifetime of a power electronic assembly. The generated data are often analyzed according to Weibull statistics. But before assuming a Weibull distribution it has to be proven that Weibull fits at all. Another problem of the Power Cycling test data analysis is the difference in test conditions between the single devices (different temperature swings), which results in a distorted number of cycles to failure. In this paper a correction method for that problem is shown, which makes it possible to use all samples of one Power Cycling run for statistical analysis.