Success Run testing - Confidence Levels

[GOZOBB]

Good afternoon,

My quality manager is asking me this question: if I conduct a success run test what confidence can I have of passing the test after 10% of the testing time, 20 % of the testing time etc ?

I was thinking of making Confidence level subject of the success run formula . However you would also need to know what is the Reliability function over time .

Am i correct in my approach?

Any feedback will be appreciated

Thanks

George

 

[Ajit Basrur]

Hi George,

Welcome to the cove

Being weekend, the response may be bit slow, so pl be patient to get responses

 

[GOZOBB]

Hi Ajit,

I perfectly understand that. Hopefully I will start getting replies on Monday.

Thanks

George

 

[Bev D]

It is helpful if you can describe exactly what you mean by a 'success run test'

also what are you testing and how?
Do you have any historical data on the reliability/stability of the product?

often the physics of the unit under test and the test itself influence the 'prediction of future success' rather than any specific statistical calculation.

 

[GOZOBB]

Hello Bev,

Thanks for your interest in my question. Actually we will be testing 18 units (this is what the testing equipment can afford) . The configuration of the units is such that we feel confident that should pass the durability test. This configuration has been chosen based on past testing and also on expert opinion.

If we pass this testing than we should be able to implement the changed configuration. The ability to predict that the test will be successful is crucial because we need to keep the trust of the eventual customer who at the moment is experiencing problems with the current configuration.


Regards

George

 

[Bev D]

sorry George - but the response you gave, while helping advance my understanding a little, doesn't put me in a better position to answer your question.

 

[Miner]

The success run approach is a form of demonstration testing based on Bayes' Theorem. This method is outside of my experience. I can help with parametric demonstration or estimation testing, but not Bayesian methods.

This thread "Confidence and Reliability - Reference to a 90/95 confidence and reliability level" may help. Here is an article from Purdue University on Bayesian techniques to Reduce Sample Size in Automotive Electronics Attribute Testing.

 

[GOZOBB]

All,

Thanks for your help. I will try to be more specific in my request by putting in numbers. So let's say I am testing 22 units for a time T with 0 failures to demonstrate a reliability goal of 0.9 with 90% confidence level.

I think it will be logical to say that once the test is started, the confidence that the reliability goal will be achieved starts increasing as that testing time increases.

Is it possible to substantiate the above argument mathematically ?

I hope I have expressed myself clearly.

Thanks in advance for any suggestions/ideas.


Sincere Regards

George