Reliability test sequence

[sabrina]

Hello,

Tarik has already given me explanations, but for a better understanding , I would like your comments on this topic .

The same parts are getting a sequence of different tests ( stress 1, stress 2...) , there are 3 dependent failure modes .

A colleague told me that I can use Blocksim if failure modes are dependent in a test sequence .

"Dependent" means that the life accumulated in a test affects the life of the part when it's undergoing the subsequent test.
( a part is tested under stress type B , then the fact that it was put under a test of type A before makes the part's life shorter)

Could you confirm if I can use Blocksim in this case and how could I calculate the total reliability ?

Please help
Thank you

[tarik]

What are you tryint to solve? how are these different stresses going to be present in real life?

[sabrina]

Hello Tarik,
I am trying to determine the total reliability of a test sequence . There is no failure

stress 1 : high temperature - 1 failure mode

stress 2 : thermal cycling - an other failure mode
....

Thank you
Best regards

[Pantelis]

So all you have are specimens undergoing an accelerated test sequence (at different stresses), with no observed failure data?

[sabrina]

Hello ,

Yes, and how could I determine the total reliability ?

Thank you for your comments

Best regards

[Ken Poi]

Dear Pantelis,
May I join in the discussion at this stage whilst not trying to detract from the discussion here. I am faced with a similar dilemma and decision as Sabrina, and I think it goes like this, How could I incorporate a series of existing tests commonly used as a design validation test matrix to determine a single reliability number? These (possibly tens) tests can be chosen to detect expected failure modes in new designs. The motive for this discussion is to integrate known Validation tests into a Designed Reliability Test(s) and then calculate that reliability. I have used Reliasoft's DRT Weibull to DRT design a reliability test with sample size n=22.7655 (23) and R=0.97 and C= 50% confidence level and F=0 (no failures and success based test). Now, I need to perform Tests sequences ( test 1, test 2, test 3)on samples group A (units 1-6), Test sequences ( test 4, test 5, ... test 8) on samples group B ( units 7-15 ) and Test Sequences (test 9,10) samples group C (units 16-23). Can I and how do I calculate the reliability of the device module A (say) after performing each individual test here. Then how do I calculate the reliability of the Module A (after all the tests selected in the Test Sequence for all the samples are completed).
Each test maybe electrical, environmental or Structural and can vary in duration for each sample group. Test Monitoring of the modules may be continous. Each test set up takes a long time to complete but the testing of each component is seldom long. Appreciatively yours
Ken

[sabrina]

Hello Pantelis, Hello Ken Poi,

Yes, it's the same dilemma .

The same parts (ECU ) are getting a test sequence with 0 failure .It's a validation plan too .

Pantelis, I'll appreciate once more to have your comments
Please help
thank you for your help
Sabrina

[sabrina]

Hello Pantelis,

I have an additional comment to make ,there is no failure .
But the failure modes are assumed dependent in case failures .

Many thanks .

Regards

[tarik]

Ken:
If you are assuming that each test type is targeting a different failure mode and that these failure modes are independent (meaning that the life accumulated by a unit on a test phase doesn’t affect the life of the unit when it’s put in a different test phase), then you can analyze the results of each Failure mode and multiply the reliability values together to obtain the reliability of the product when it’s being used in the field under all types of stress.

However if you can’t make that assumption, then read my comment to Sabrina:

Sabrina:
Based on what you have described so far about the way you conduct your test, and if you can’t make the assumption of independence, then currently ReliaSoft tools can’t be used to treat your data, because the test wasn’t properly designed. The test you are conducting doesn’t adequately reflect the real world application. In the ‘real world’, a unit doesn’t go through a sequence of stresses, they are usually present at all times, or happen randomly. Unless your test actually reflects your real operating conditions, meaning in the real world, your units will go through stress1, for a certain period of time, then stress2 , for a certain period of time..etc. In that case, your test has been properly constructed. And in that case I would consider these stress phases as ‘one’ type of stresses (consider it ‘real world’ stress) and I would just record the failure times and analyze them.

It would have been better to make all your units undergo a test that includes all the types of stress simultaneously. If you don’t have a way to combine your stresses in one test, you could test different units under different stress types and obtain reliability values of each test condition (failure mode) and multiply the results together to obtain a ‘general’ reliability value for the product when it is used in the real world

[Ken Poi]

Tarik, Sounds like the better way to determine the reliability of the modules I want to test is to create a Single test called a Reliability Test which will run for a fixed time subjecting all the DUTs to the same stresses or test sequences in one go and then compile the failure data (mixed failure modes, failure times ) and plot it with a Mixed Weibull and get the statistics out from that plot. A binomial distribution can be assumed for the sample size determination which is what Weibull++ DRT calculates easily.

Now the challenge lies here where a single comprehensive system level test that exercises these subsystem components will actually in one go combine all the electrical, climatic and structural stresses we intend to subject the modules to has to be designed, a all in one qualitative reliability test. So far as I can read from this forum this has yet to be achieved yet as the present designed tests are really characteristic type tests on the DUT modules under test. My question therefore has it been accomplished in the electronics industry and how was it reliability tested.