The
mathematical models to represent the relation- ship between the
thermal stress and the deterioration rate for electrical insulation
are well established based on the Arrhenius law and appropriate
underlying probability distributions for constant thermal stress.
There are two kinds of methods to deal with the thermal lifetime:
one is the two-valued discrete model representing 0 for alive and
1 for dead, and the other is to use continuous value for the mechanical
strength deterioration of the material. In this paper, we deal
with the latter case. In the IEC (International Electrotechnical
Commission) 60216-1, deterioration due to the thermal stress is
represented by the mechanical strength, and the time showing 50%
mechanical strength to the initial strength is defined as the failure
time. For underlying probability distribution models, the normal
dis- tribution, the generalized Pareto distribution, and the general-
ized logistic distribution models are proposed. Based on these
mathematical models proposed, we investigate the optimum life test
plans for the 50% mechanical strength index. Using a real experimental
case, we obtain the adequate parameter values for the simulation
accelerated life test. The optimum parameters we seek are the most
appropriate number of specimens at each stress level. Comparing
the optimum test results and the conventional test results by equally
allocated to each stress, the ratios of the prediction error for
the optimum tests to that of conventional tests are around 85%.
This consequence is very similar to that in the case of two-valued
discrete model. |
|
|
|
|
blood
type, personality, new questionnaire, human intuition, Bayes
theory, imprinting.
|
|
|