Statistical analysis of Truncated-Data Method to Shorten Thermal-Aging Tests of Electrical Insulation

Abstract : The standard tests of insulation materials (motorette tests and magnet-wire twist tests) call for aging ten specimens until failure at each of three or four temperatures to obtain a life-versus-temperature curve. Valuable time could be saved if an estimate of the life at each temperature is valid from data truncated at fewer than all ten failures. Therefore a statistical analysis was made, using 50 previous thermal evaluations, of the validity of three truncated-data estimates of the life at each temperature: the log average of the times for the fifth and sixth failures (method A), an estimate from a probability-plot fit to the first five failure times (method B), and simply the fifth failure time (method C).