Is Failure Rate Constant for a Complex System?


McLinn, James A.   (1989, ASQC)   McLinn, Dale and Associates, Minneapolis, MN

Annual Quality Congress, Toronto, Ontario, Canada    Vol. 43    No. 0
QICID: 3618    May 1989    pp. 538-545
List $10.00
Member $5.00

This article is not available online. Contact us to receive a scan of the archive, in PDF format.

Article Abstract

A large, complex and expensive software controlled system was developed and deployed. The system consisted of three subsystems; an "AT" type single card computer with associated control electronics, precision three-dimensional movements, and fluidics components with associated control. During the first year of operation in the field, reliability was carefully monitored. Field performance, reliability growth as well as improvement efforts were all monitored for effects on performance. Installation, maintenance and upgrade data were all available.

This system with its complex nature and high degree of documentation could answer the following questions:

  1. Determine if a bathtub curve is appropriate.
  2. Check the assumption for the system as a whole or any subsystem of a constant failure rate during the useful life.
  3. Look at Duane Growth and the effects of quality improvement programs and corrective actions on a manufacturing lot by lot basis.
  4. Estimate the life of components and subassemblies.
  5. Check the actual field failure history against the prediction models in Mil HDBK 217 and Bellcore.
This paper will check the hypothesis that a complex system will follow the bathtub curve and have a constant failure rate after the infant period. All the subsystems will be separated and checked independently. System operation was at room temperature, benign conditions. Wearout of some components was observed and so the effects of periodic maintenance could be observed.

Weibull graphical analysis techniques were used to analyze the data from the field and this was compared to a small in house population of systems. Approximately 50 field systems scattered around the U.S. and Canada serve as a stable population for estimating failure rates. Raw data from the field and in house would support a Weibull slope of 0.8 and 1.0 would be a constant failure rate.


Reliability,Tool wear

Browse QIC Articles Chronologically:     Previous Article     Next Article

New Search

Featured advertisers

ASQ is a global community of people passionate about quality, who use the tools, their ideas and expertise to make our world work better. ASQ: The Global Voice of Quality.