Why chosing p=1.5 (accelerating power for the mechanical stress)? What is the reference/rationale for the p value?
P.S. Dave Steinberg′s "Vibration Analysis of Electronic Equipment" suggests p=6.4 for solder joint fatigue.
Why chosing p=1.5 (accelerating power for mechanical stress)
Wed, 06/04/2011 - 16:59
#1
Why chosing p=1.5 (accelerating power for mechanical stress)
Edited by: Lambda on 28/08/2012 - 10:18
FIDES Guide 2009 Edition A
Section 2.7.1 Physics of failures and models (page 44)
Why chosing p=1.5 (accelerating power for the mechanical stress)? What is the reference/rationale for the p value?
P.S. Dave Steinberg′s Book "Vibration Analysis of Electronic Equipment" suggests p=6.4 for solder joint fatigue.
Thank you for your question and your close reading. Dave Steinberg′s book is one of our main references!
Indeed, the paragraph 2.7.1 page 44 of the FIDES Guide is largely meant to answer this question and the end of this paragraph is to warn about some consequences of this power law. To be simple, solder joint fatigue is not the only failure mode covered by this law.
Furthermore, the theory of physics of failure considers a strong causality between the physical phenomenon and the failure. In reliability prediction, we assume that the relation is more complex. The 6.4 power will be true for laboratory trials, looking only at SnPb solder joint. On the field, our point of view is that a power of 1.5 is more realistic.
The calibration of this factor has been done during the development of the first edition of the FIDES Guide, by comparing the accuracy of several reliability predictions in various environment. The 1.5 value has been confirmed by the studies and experimentations done to validate the 2009 edition.
I hope this helps!
All the best with FIDES,
Lambda
Thanks for your info.
For Section 2.7.1 "Physics of Failures and Models" (page 44)
Would the Grms be for individual part, PCB, or CCA?
If I understand well your question, the answer is in the chapter 2.7.4.
It is stated that the model parameter setting will be made using the input level for the product (the level experienced by the equipment or the board).
Lambda
Cher M. Lambda,I just read that the preferred language is English - a great relief for me. I have a related question about the p-value to be used. I have to do an accelerated life test for galley equipment. One of the stressors to be considered is vibration stress. For this I want to use Basquin's law, as is also recommended by FIDES. From FIDES I learn that if I want to include multiple failure modes in my testing I should use p=1.5 in my calculation, so the most conservative value for all failure modes. This was also explained to more detail in this forum; so far I everything is clear.However, when I look at M. Giraudeau's presentation at RAMS 2011 (slide 18), I read that "The value of 1.5 is relevant for reliability prediction, not to accelerate testing." I think that is contains additional information, but it confuses me. I would guess that the lowest p value requires the longest time time, so I should use p=1.5 anyway. Only in the case I have a failure I think I should subsequently determine its p value by repeating the experiment at different vibration levels. Then I can use for this failure mode the freshly determined value. Am I overlooking something here? Did I misunderstand something completely? Any help would be greatly appreciated.
Dear,
Indeed, the point you quoted from M.Giraudeau's presentation is the shared and common point of view within the team that developped FIDES. I'm not expert in environement enginnering but to my knowledge accelerated testing are never designed with the acceleration factor of 1.5 used in FIDES. The reason why reliability previsions don't fit well with field return if a stronger acceleration factor (like the one used for lab tests) is used is quite complex and can not be summarized here (nevertheless, see page 44 of the FIDES Guide).
I recommand you to take advice from experts in the field of environement engineering, FIDES is not the right guide for testing and trials.
Regards,
Lambda
Dear Lambda,Thank you for your prompt reply. I took me some time to get some feeling why p-values for testing and reliability calculations can be different from each other, but I think now that this must be due to the fact that upon testing one has the opportunity to do a failure analysis if a failure should occur, while for calcuations one cannot. So in testing one can distinguish between failure modes, while for calculation one cannot, right? Can you recommend some literature reference (book, paper, etc.) on this topic?I do have one other question, I'll post it in the relevant topic. Many thanks again.