|
|
-
Guaranteed
Long-term study by statistics' William Meeker looks at product reliability.
-
Customers demand high reliability in products. That's a given regardless
if the consumer is purchasing a new automobile, personal computer or toaster.
William Meeker, Distinguished Professor of liberal arts and sciences and
professor of statistics, is no exception to that rule.
"Product reliability is extremely important whenever I buy something,"”
he says. "Consumers have come to expect that reliability needs to
be built into the product design. A 'ship and fix later' approach is no
longer acceptable."
Meeker knows this first-hand more than the average consumer. Not because
his refrigerator, cell phone or lawn mower has quit working.
No, Meeker's knowledge of product reliability is the result of a long-term
study he has undertaken on methods of speedier product reliability analysis.
"I've had exposure to these kinds of problems for a number of years,"
said Meeker, who researched some of these efforts during summers at Bell
Laboratories from 1978-92.
Meeker's studies, which he has conducted along with statistics colleagues
at General Electric's Global Research Center, Gerald Hahn and Necip Doganaksoy,
have looked at using accelerated life tests to predict product reliability
and in some cases, warranty returns.
Accelerated life tests provide statistical assurance that reliability
goals can be met and an early warning if they cannot, Meeker says. There
are several types of accelerated testing including use-rate acceleration,
product aging acceleration and product stress acceleration.
Suppose that a product, such as a toaster, is used only once a day in
the home and that it can be tested 365 times in a laboratory on a single
day. If no problems arise in five days of testing, then the toaster life
span should be at least five years.
There are difficulties with this type of product testing according to
Meeker.
"What happens in a controlled situation like a laboratory testing
facility and what happens to that same product in the field are two different
things," Meeker said.
He says interpretation of product accelerated life test results should
take into account the laboratory results, field usage patterns and other
environmental factors.
"Once industry finds out how people use their product, they can connect
those results with what happens in the lab," he said. "The closer
the assumed model is to the physical situation, the better.
"Accelerated life tests can provide statistical assurance a company's
reliability goals can be met with regard to the accelerated failure modes
and provide an early warning if they cannot," he said.
Meeker, who has worked with General Electric, Hewlett-Packard and 3M on
these types of statistical models, has also collected his own personal
data on products such as light bulbs and automobile tires.
He has also looked at various warranties, especially those provided by
automobile manufacturers. While he says companies use warranties as a
marketing tool, they also want to make sure their product is highly reliable
for at least the length of the warranty.
"Otherwise if a problem does show up it will have a serious negative
impact on the bottom line," he said.
Around LAS
October 18-31, 2004
|
|
