 Back to Programs
Sample Size Conundrum for Reliability Testing: How Big
Or Small ?
Nihar Senapati,Lead Reliability Engineer
GE Healthcare
439 South Union Street 3rd Floor,Lawrence,MA 01843
Nihar.Senapati@ge.com
Reliability practitioners and design assurance engineers tend to confront a dilemma
while selecting a sample size. In most cases the size of the sample is arbitrarily
chosen convenient to the situation and availability of the sample test units. Limited
understanding of the design assurance engineer or reliability test personnel on
sample size estimation process dilutes the primary concept and spirit of reliability
test process. Small sample size produces test results, which don’t buy the approval
from management decision makers. Confidence in such results is usually very low.
On the other hand, a more than adequate sample size (though augurs well on the meaningfulness
of test conclusions) tends to be cost and cycle time prohibitive. Sample size estimation
process can be a challenge starting from a simple reliability demonstration test
to conducting a Highly Accelerated Life Test (HALT).
The paper focuses on the several risk factors including producer’s risk and consumer’s
risk and various cost factors while discussing an optimal sample size. This paper
also discusses sample size with consideration to various apriori statistical failure
distributions. The goal is to showcase several paths for sample size determination
in the industrial environment and reduce errors in reliability testing.
Presenter:
Nihar Senapati Nihar Senapati is a Lead Reliability Engineer with GE Healthcare in Lawrence, MA. He works in hardware and software Reliability areas and holds a MS in Reliability and Quality engineering from University of Arizona, Tucson.
He is also a CRE, CQE and a Six Sigma Black Belt from ASQ. He has consulted and published many papers on various aspects of Reliability, Quality and Six sigma process improvement initiatives. Nihar is a member of ASQ and IEEE Boston Reliability Chapter.
Back to Programs
|
