Accelerating Life testing using Thermal Cycling

[Greg Spraker]

We currently have a procedure that tests a device at an elevated temperature for 6 months (using the acceleration factor based on arrhenius). Is there anyway to accelerate this test using thermal cycling - if so, what distribution can I use to determine reliability and confidence intervals?
Thanks in advance for any help,
Greg

[Pantelis]

In an accelerated test one uses a stimulus to accelerate the failure mechanisms under consideration. In your case you mention temperature as that stimulus. The data obtained from the test under that stimulus can then be analyzed with an appropriate life distribution (i.e. Weibull) and a life-stress relationship (e.g. Arrhenius, etc.) {See http://www.weibull.com/AccelTestWeb/arrhenius_weibull.htm for a discussion on this combination}. From that analysis one can then determine the items you seek. This can be applied when the stimulus is constant. Now, under cycling if the assumption is that the stimulus is only the temperature change (i.e. rate of change is not a stimulus) the same methodology can be used ( as an example see http://www.weibull.com/AccelTestWeb/ipl_and_coffin_manson_relationship.htm ). Otherwise one may want to model the rate as a second stimulus, etc.

You may also want to see some examples at: http://www.reliasoft.com/alta6/a6examples.htm

I hope this helps

[Jason]

Is there a general set of parameters (change in temp, rate of temp change) that one could use for an accelerated test using cycling? for example, how far apart the high and low temperatures are; rate of change of temperature; soak time?

[Pantelis]

These values are product specific ...

[RS Support]

You may also want to see a recent article in at http://www.weibull.com/hotwire/issue19/hottopics19.htm