SRC Forum - Message Replies
Forum: Reliability & Maintainability Questions and Answers
Topic: Reliability & Maintainability Questions and Answers
Topic Posted by: Reliability & Maintainability Forum
Organization: System Reliability Center
Date Posted: Mon Aug 31 12:47:36 US/Eastern 1998
Posted by: Ken poi
Organization:Australian Arrow Pty Ltd
Date posted: Mon Jan 10 9:30:55 US/Eastern 2005
Subject: Reliability in the Automotives
I am setting up an ORT SOP for an Body electronics module of a truck. I am advised to use current test matrices used for Design Validation. I understand the failure modes this type of electronic device are usually solder wetting related. I am contemplating sampling 12 units (arbitrarily chosen) for a barrage of tests which are targetted at precipitating the failure modes mentioned above.
My goal is 0.97 reliability (meaning p=0.03) with 50% Confidence Level. ( There is no table for p=0.03 in your toolbox, Can it be generated ? )
I propose that some of these 12 units will be run through a series of destructive tests and the others through another series of tests like Power Temp Cycling, Vibration testing, Altitude testing and Temp Storage tests.
Test Name 1-3 +1 extra 4-6 7-9 +1 extra 10-12
HIGH TEMPERATURE - STORAGE * *
LOW TEMPERATURE - STORAGE * *
HIGH TEMPERATURE - OPERATION * *
LOW TEMPERATURE - OPERATION * *
RAPID CHANGE IN TEMPERATURE * *
TEMPERATURE CYCLING * *
DAMP HEAT STEADY STATE * *
TEMPERATURE STEPS * *
THERMAL CYCLING & VIBRATION COMBINATION * *
COMPOSITE TEMPERATURE / HUMIDITY CYCLIC * *
Objective is to demonstrate reliability of the product.
Q: Can I validly use the binomial distribution assumption to define sample sizes and proportion defective in the sample for these samples assigned for their series of tests? Assume no failures will be accepted. Using the tables available from the toolbox on this website with assumptions of proportion defective P=0.10 and expecting no failures, the sample size of 12 ranks in between a confidence levels of 60%-80%. Not satisfactory enough.
Q: Does this result suggest that to attain a CL of 90% and expect no failures from the sample, I have to increase the sample size from 12 to 22?
Q: Are my assumptions this type of series of destructive testing (or trials) each has an equal probability of success thus a binomial distribution can be validly applied to its sample determination with a fixed number of failures.
Subject: Confidence Levels and Sample Sizes
Reply Posted by: bwd
Date Posted: Thu Jan 13 9:53:09 US/Eastern 2005
We reviewed your lengthy test plan and the overall determinations are: one cannot run a binomial destructive test, the sample size for a 90/90 test will require at least 22 units, and a test for 97/50 would also require 22 units. Details of our evaluation are enclosed as an attachment sent directly to you and include the 0.03 probability factors.