Principal Engineer - Xerox Corporation
TG Rainbow Incorporated
PE, CRE, CSSB and Six Sigma Master Black Belt - ASQ
As an American Society of Quality Six Sigma instructor and mentor, I have observed abundant evidence that the effective utilization of reliability specifications can be a challenge to
engineers. To address this challenge, I propose for your consideration a simple step-by-step approach to implement and integrate reliability specifications:
- Benchmark current reliability numbers for the function you are replacing in your new design.
- Make sure that you consider both your past products and the products of your competitors.
- Understand the current failure modes of the best product and innovate to improve the reliability.
- Ensure that you do not create new failure modes.
- Based on the leading failure mechanisms of the system or component, determine the survival distribution by testing. I recommend accelerated testing.
- Specify the 2%, 50%, 63.2% (Characteristic Life) and 98% failure points in hours, days, cycles or years, as applicable.
- Ensure that you call out the real stresses under which the component or subsystem is used. This is critical for getting good correlation of test data to real life.
- Determine critical parameters that drive the quality and reliability.
- Specify the reliability and critical specification of parameters that drive the failure mechanisms.
- Parameter specification comes from characterization and optimization test results.
- Include a warranty note in the drawing to enable proper controls at the supplier for the quality today, tomorrow and thereafter.
- Ensure that proper field returns procedures are in place to find root cause of failures.
- Ensure that you have resources and tools available for immediate feedback to the source of problems.
To demonstrate the effectiveness of this methodology, I would like to conclude with an example of its implementation. The example involves a clean sheet design of a complex product intended to
replace the existing fleet with lower cost units at no extra service cost. When the activity started, the actual was 1 million cycles and the target was 5 million cycles. By implementing the
methodology described above, the company was able to achieve a design with predicted reliability from accelerated testing of 26 million cycles. For one component, we were able to improve the
reliability by 26 times at no extra cost to the component, two years before launch!