Transference of Test Results from Experiments on Small Models to N Times Larger Test Objects with Insulation under Oil

Abstract

To dimension large transformers optimally, breakdown tests are carried out on simplified models using a 1 X 5-μsec (microsecond) positive polarity wave. Tests show that the distribution of breakdown voltages of a random sample is Gaussian. Assuming a basic Gaussian distribution, the validity of the area effect can be demonstrated unambiguously. The hypothesis that results obtained by repeatedly stressing a single model n times can be transferred to an object n times larger, cannot be confirmed by the tests.