Page 274 - Handbook of Materials Failure Analysis

270 CHAPTER 11 The reliability design of mechanical system

The cumulative failure rate of 2% in the expression B2 life requires that we should
test at least the inverse of 2% or 50 units. Two percent means two failures in 100 test
items or one failure in 50 items. To attain the target B2 life 5500 h, we should test
50 units for 1 month under the above conditions. If no failure occurs, then the driving
module meets the target with a common sense level of confidence. If we want to reduce
the sample size, the test time must be increased. Doubling the test time to 2 months
reduces the sample size one-quarter (square of the inverse of the test-time multiplier),
or 13. The final test specification, then, is that 13 units should be tested for 2 months
under elevated load and temperature conditions with the criterion that no failure is
found. (To attain the overstress failure target of 1% per 5 years.)
With this simple procedure, it is possible for any engineer or stakeholder, even
the CEO, to mentally calculate the sample size and testing time (based on target life-
time, AF, and test-time multiplier). All key personnel would be able to raise simple
questions about test credibility, such as the validity of the sample size, test duration,
AF, and so on, making corrections obvious and the new product more reliable. After-
ward, it is enough for CEOs to check the metrics or reliability directly related to the
outcomes of their businesses—the lifetime target and the annual failure rate.
Note that with this method, we can obtain the reliability state of the production
lot, but we do not know the time when and the site where each product will fail. This
is because each item may have been made of a somewhat different material and by a
slightly imperfect process. Furthermore, it is very difficult to estimate the next failure
in item use because minute structural flaws caused by different materials and pro-
cesses may have been magnified during use by various stresses that cannot be exactly
identified. But it is not difficult to extend the lifetime of the production lot as an
aggregate because we could test it longer, identify the failure site, and improve it
with related correctives. Thus, it is easier to lengthen the lifetime of a certain item
before its release into the market than to extend the lifetime of an item being used in
the field.

6 PARAMETRIC ACCELERATED LIFE TESTING
Parametric ALT for mechanical system like automobile in Figure 11.8a is an accel-
erated life testing to find the optimal design parameters not considering in design.
First of all, it is critical to formulate the accelerated factor from the life-stress
model and the dominant failure mechanism under the given physical and chemical
conditions. Knowledge of physical of failure is required to understand the failure
mechanism—traditionally fatigue or fracture in mechanical system due to repetitive
stresses—related to a life-stress model.
Parametric ALT for RQ involves three key steps:
1. Create a life-stress model, determine the accelerating factor under severe
conditions, and calculate the minimum testing periods required by dividing the
target lifetime by the AF.