Determining the duration of a demonstration life-test before all units fail

Two small samples of electrodes insulated with the standard and modified designs were put on a high-stress voltage-endurance life-test. The objective was to compare the life distributions of these insulation designs. During the test, the design engineer suspected that the modified design was an improvement (longer lasting) over the standard design. There were still unfailed electrodes in the modified sample when all the electrodes in the standard sample had failed. However, a statistically significant difference-between the α (scale) parameters of the assumed Weibull distributions for insulation life-could not be shown at the accumulated testing time. The design engineer wished to know how much longer the unfailed electrodes in the modified sample needed to survive in order to provide convincing evidence for the difference between the two designs. This paper presents a procedure to address this question and discusses further aspects of determining test duration in similar practical situations