Variance reduction in aged sample data by the use of lot acceptance test data

摘要

NAVSEA Crane Division performs ongoing Quality Evaluation Test (QET) programs to evaluate the steady deterioration with age of critical performance parameters (such as risetime and runtime) of Navy Fleet-aged missile primary batteries. The goal is to estimate service life (SL), the age at which these batteries must be replaced. QET data is. obtained by sampling a few dozen units of various ages throughout the range of date of manufacturing (DOM) of the inventory. It is expensive to obtain; few data points are available. Typically, a QET is performed every 3-5 years on 22 samples. Abundant Lot Acceptance Test (LAT) data usually is available; it is used to perform lot sentencing (accept or reject) before shipping. Here we show new mathematical methods which improve the accuracy of service life estimation, and which also may be a significant advance in the general mathematical theory of regression analysis. The Enhanced Delta Method also eliminates the well-known problem that arises when QET and LAT raw data sets are combined in one analysis.