Newly developed flashover voltage test methods
are proposed in this paper. These methods estimate the mean value
of the distribution efficiently and stably using breakdown and
nonbreakdown responses when the flashover voltages follow a normal
distribution with known coefficient of variation. Electrical engineers
have long been using the up-and-down method developed by Dixon & Mood
(J. Amer. Statistical Assoc. vol. 43, pp. 109-126, 1948) to assess
the flashover voltage distribution when both the mean and the standard
deviation are unknown. However, the method is numerically unstable
in obtaining the mean value when the sample size is extremely small,
e.g. 5 to 10. The proposed methods, the "mean-shooting method" and
the "modified up-and-down method", are extremely effective
in a numerical stability sense compared to the ordinary up-and-down
method when the mean is unknown but the coefficient of variation
is assumed to be known in advance. As for the estimation error
of the mean value of the flashover voltage, the error obtained
by the mean-shooting method is seen to be superior to that obtained
by the ordinary up-and-down method; for instance, 15 shots in the
mean-shooting method is equivalent to 20 shots in the up-and-down
method.
|