Setting up and analyzing a two stress accelerated test

Planning a viable two-stress Accelerated Test can be a good challenge. Determining the stresses is just the start of the reliability challenge. It starts with understanding any relevant history and the associated failure modes. The root cause(s) of the main field failures would represent a good set of stresses. Some customers may experience as many as five simultaneous operating stresses in the field, yet only two or three might be determined to be major. A common stress combination is temperature and vibration, yet sometimes a better combination might be temperature and humidity. Mobile systems might need mechanical loads combined with temperature extremes to best represent the field. Systems exposed to sea air might use salt air combined with temperature. Functional test parameters with degradation measures may be required to obtain meaningful test results of long lived products. Difficulty in handling the stress combination combined with the possibility of non-linear behavior may result in changes to an accelerated test. Add intermittent or soft system failures to this mix and reliability challenges increase. Data analysis of a small number (i.e. two or three samples) increases difficulty of analysis. Sample size selection should represent as wide a variability as possible. Often available samples is smaller than desired and this complicates test planning and results. Degradation measures become indispensable when zero failures occur in test. Data collection times during test may also impact the analysis especially when looking for non-linear behavior. Test data collection points are set for convenience of reading and not to yield the best spread of information for analysis. This paper will present several detailed examples, covering the best methods for selecting stresses for accelerated testing and implementing the test. These examples show practical sample size and test time collection points. Data handling issues to clarify results even when noise in measurements is present will be discusses. Lastly, a short discussion of analysis. Planning for a myriad of possible results should help prevent unexpected events that damage ability to understand the results.